| blob | 9675c322134cec66e4de6909b613b799ba2f87e9 |
1 2011-02-21 Andreas J. Koenig <andk@cpan.org>
3 * rrr-server has a memory leak
5 root 16063 94.7 10.6 358404 354600 pts/21 RN Feb18 4117:16 /home/src/perl/repoperls/installed-perls/perl/v5.13.8-16-gf53580f/bin/perl -Ilib bin/rrr-server --verbose /home/ftp/incoming/RECENT-1h.yaml
7 2011-02-20 Andreas J. Koenig <andk@cpan.org>
9 * start consistently considering the RECENT files themselves not part of
10 the set.
12 2011-02-19 Andreas J. Koenig <andk@cpan.org>
14 * Found the term anonymous one way file system
16 * No support for empty directories
18 2011-02-17 Andreas J. Koenig <andk@cpan.org>
20 * .recent/ or .rrr/ for metadata
22 * ALERT rrr: we have a piece of code somewhere in the mirror()
23 subroutine that reads the raw YAML file with normal open() and cuts it
24 off for efficiency, then feeds it to yaml::load in order to get some
25 metadata. Reading the thing with the whole tail of the recent array
26 kills the whole idea. The whole thing will not work with JSON.
28 So we must invent a protocol 2 that works with one additional index
29 files, say "A", only containing meta.
31 2011-02-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
33 * I think interval guaranteed with the $last_aggregate_call
34 variable (both in fsck and server) need to be fixed with a $recc
35 variable to something like 60 seconds or even less.
37 * minor bug: fsck just added the lockfile to the index which should be
38 considered bookkeeping.
40 * speed: a large directory remove that is done by the kernel in 1
41 seconds triggers minutes of bookkeeping (07:56:42 - 08:01:10 for ~3000
42 files, all due to Schlemiehl again. Want to do some collecting of
43 pending ops before actually writing to the RECENT files. Lock!
45 Repeating the timing with 2600 files and it took 21:28:08 - 21:35:27.
47 Now having rewritten the loop to use batch_update(): 4383 file removed
48 21:43:37 - 21:44:12. From 7:30 to 0:35 while doing 60% more work that's
49 a joy.
51 * lockdirectory expiration? server died and blocked fsck for so long.
53 * At the moment rrr-init is not needed anymore, rrr-server can be
54 started and followed by rrr-fsck. Still stupid but not the main
55 showstopper.
57 2011-02-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
59 * IN_Q_OVERFLOW needs to trigger an fsck, likewise the entry into the
60 server.
62 2011-01-30 Andreas J. Koenig <andk@cpan.org>
64 * rrr-init and rrr-server need to be used such:
66 init
67 server
68 init again
70 Stupid but no showstopper ATM.
72 2010-10-26 Andreas J. Koenig <andk@cpan.org>
74 * forking slave is done, needs more testing.
76 2010-07-06 Andreas J. Koenig <andk@cpan.org>
78 * The two next most important goals are to get a useable rrr-server and
79 a forking slave implementation that uses virtually no memory after the
80 first pass through. Does it need the status file?
82 * rrr-init now being fast we can now remove the alpha alerts. Done.
84 2010-06-16 Andreas J. Koenig <andk@cpan.org>
86 * rrr-init seems to be sssslllloooowwww. At 3:49 I started it for the
87 video tree. After twelve minutes the file has reached a size of 500k. It
88 is locking and rewriting the YAML after every file. Wasn't there an
89 unlocked variant? For "init" one would not expect locked correctness and
90 a rewrite after every item. Given that this tends to become slower with
91 every file it is really unacceptable.
93 Reminds me of rrr-dirtyupdate (the former name of rrr-update). It seems
94 the code was not dirty enough and that was the reason why it was so
95 slow. And rrr-update does not have a --dirty switch, so it looks like we
96 have no way to force dirty and fast operation yet but we need it for
97 init.
99 Irritating the bootstrapping race: the order {(1) fill initial
100 array, (2) start server} leaves the need for an fsck to cover the time
101 between 1 and 2. Given that regular fsck is needed in any case the price
102 is not high but still doesn't give a warm feeling. Maybe starting the
103 server will by itself do an fsck?
105 2010-06-15 Andreas J. Koenig <andk@cpan.org>
107 * remove the wild alpha alerts and say, it is alpha but parts of it are
108 already working very well or so.
110 2009-12-10 Andreas J. Koenig <andk@cpan.org>
112 * Prominent request for statusfile to safe memory consumption. I'd like
113 to work backwards: get a verification program that verifies how well in
114 sync two nodes are without using frmr itself. Pretend that the forking
115 agent already works and can be interrupted with ^C. Then let it die at
116 random and restart at random and see the memory consumption and the
117 failure rate over time. Compare it with the curve we have now.
119 Of course we also invented the status file as a sharp debugging tool.
120 I'd like to see it in action. See also: F:R:M:Recent::thaw().
122 2009-06-22 Andreas J. Koenig <andk@cpan.org>
124 * Todo: rrr-init to generate some index files so that rrr-server can be
125 started.
127 2009-05-02 Andreas J. Koenig <andk@cpan.org>
129 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
130 so we can use it for normal update with a single file. And add a cronjob
131 that runs rrr-fsck so we get a reminder when something is not working
132 as it should.
134 2009-04-27 Andreas J. Koenig <andk@cpan.org>
136 * Todo: I still see more link_stat errors than expected. I come back to
137 the former suspicion that we do not do necessary bookkeeping on deleted
138 files so that when we reach the "new" event on a meanwhile deleted file
139 we try to mirror it although we could know better. It's just noise, no
140 harm but noise is irritating.
142 * Todo: more experimenting with runstatusfile to get it official
144 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
145 parse the error.
147 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
149 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
151 And that ignore_link_stat_error is propagated in sparse_clone.
153 DONE
155 For the record: the bug was that ignore_link_stat_error was lost during
156 sparse_clone. And then the second bug was that it did not default to
157 true. Both are now fixed.
159 2009-04-25 Andreas J. Koenig <andk@cpan.org>
161 * Todo: fill rrr program with life
163 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
165 * Minor bug: I see complaints about files not existing like
166 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
167 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
169 In the RECENT files these are delete events but of course also new
170 events in older recent files.
172 I think we do not properly remember deletes when we skip-deletes and so
173 do not filter them out when we later see the "new" event.
175 Yes, but that what ignore_link_stat_error is for, we know about this
176 race condition. We must make ignore_link_stat_error default to true.
178 DONE
180 * bug in Done.pm failing to merge two fields into one when a third is
181 present. FIXED and accompanied with new test.
183 2009-04-24 Andreas J. Koenig <andk@cpan.org>
185 * Todo: keep the index clean and run some kind of nondestructive fsck 4
186 times a day.
188 * who is our backbone?
190 http://cpan.cpantesters.org/authors/02STAMP # barbie
191 http://cpan.solfo.com/authors/02STAMP # abh
192 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
194 2009-04-17 Andreas J. Koenig <andk@cpan.org>
196 * the changelog helper I included in the Makefile release memo does not
197 show tags and is much less useful than I thought. gitk is probably much
198 more convenient.
200 2009-04-16 Andreas J. Koenig <andk@cpan.org>
202 * ABH writes:
204 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
205 #### Ydr_.yaml ... DONE
206 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
207 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
208 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
209 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
210 #### gone already at cpan-pause.pl line 0
212 Running without skip-deletes now but cannot reproduce.
214 2009-04-15 Andreas J. Koenig <andk@cpan.org>
216 * consider state directory such we can restart after a ^C or come again
217 from a cronjob.
219 Reuse _runstatusfile: write atomically, see how we can use the dump to
220 start running again. Do not leak and remove globs before writing. Do not
221 only dump one rf, dump the whole complex.
223 * consider the effect of ^C. Is it safe?
225 * want to use Perl::Version to up the versions on every release. Only if
226 things have changed, of course. Does this actually work?
228 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
229 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
231 Must do the bookkeeping myself when this should happen, maybe based on git?
233 I think I prefer the same version number for all pm files. But one also
234 needs a routine for setting them. Ahh:
236 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
238 NEEDSMOREWORK
240 * why indexer wrong?
242 cpan[2]> m /Mirror::Recent/
243 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
244 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
245 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
246 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
247 4 items found
249 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
251 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
253 I just tried MM 6.5102 and it again produces only a provides for Recent,
254 not for the other modules.
256 My own fault. FIXED in the Makefile.PL.
258 2009-04-13 Andreas J. Koenig <andk@cpan.org>
260 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
262 * is it true (as stated at
263 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
264 that rsync can lead to errors when upstream changes before client sync
265 has completed? Can we see the error when we run this on upstream:
267 % perl -e '
268 use Time::HiRes qw(time);
269 while (){
270 open my $fh, ">", "changingfile.txt.new" or die;
271 print $fh (time."\n") x 1000000;
272 close $fh;
273 rename "changingfile.txt.new", "changingfile.txt" or die;
274 }
275 '
277 And on the receiving end:
279 % while true; do
280 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
281 sleep 1;
282 done
284 I cannot get it to fail. Apparently it is sufficient when upstream
285 always writes atomically (which of course is mandatory)?
287 SENT to mailing list Apr 28.
289 * from ABH:
291 > https://fedorahosted.org/InstantMirror/
292 > https://www.redhat.com/mailman/listinfo/instantmirror-list
293 > irc.freenode.net channel #instantmirror
295 A cool name for a project. Inspires me to write a few sentences about
296 bittorrent's role in the grand picture.
298 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
299 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
301 2009-04-12 Andreas J. Koenig <andk@cpan.org>
303 * Interesting last minute bug during real download testing: the output
304 isn't as pretty anymore, it seems that more work is being done than
305 needed.
307 Introducing a highlevel _alluptodate method helped a lot in debugging
308 this.
310 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
311 different on 1h file than on the other files. This lead to frequent
312 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
313 issue.
315 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
317 * Todo: try if we can be faster with a native float package. I'm really
318 glad to have a machine exposing perl floating point bugs, it forced me
319 to retract this item from the changes file: "several speedups in the
320 fakebigfloat code" before I get cpan testers fails.
322 2009-04-11 Andreas J. Koenig <andk@cpan.org>
324 * Bug: when the dirtymark changes upstream, then the downstream server
325 notices it immediately and mirrors the "1h" file but then leaves the
326 rmirror loop and continues with "6h" on the next round through. This
327 behaviour goes on until it reaches the Z file. Because we eat the Z file
328 piecemeal we leave the loop after a while but after that everybody seems
329 to have forgotten that there is some work left to be done. Bad, bad,
330 bad.
332 ETOOCONFUSING better name?
334 Recentfile::get_remote_recentfile_as_tempfile OK
335 Recentfile::resolve_recentfilename split_rfilename DONE
337 Recent::_principal_recentfile_object _principal_recentfile_fromremote
338 Recent::principal_recentfile OK
339 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
341 some void, some not, some not void but called in void context.
343 2009-04-10 Andreas J. Koenig <andk@cpan.org>
345 * Bug? When the dirtymark changes then the second tier hosts quickly
346 reset their DONE state and restart mirroring. But then they mirror
347 potentially outdated index files with inconsistent dirtymark because the
348 first tier box may be a bit behind on the large recentfiles.
350 This means we need a brake when iterating over the recentfiles that
351 refuses to use a recentfile with the wrong dirtymark.
353 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
355 * install the bin/* files and talk about rrr-overview somewhere.
357 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
358 created outside the target tree.
360 * Barbie asks for better logging capabilities
362 2009-03-29 Andreas J. Koenig <andk@cpan.org>
364 * known bug: cannot configure the Recentfile objects to keep delete
365 events. How to call the accessor for that? keep_delete_objects_forever?
367 Sounds acceptable. FIXED (but untested).
369 * known bug: keeps temporary index files lying around.
371 2009-03-24 Andreas J. Koenig <andk@cpan.org>
373 * Some equivalent for
375 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
377 ? DONE
379 2009-03-22 Andreas J. Koenig <andk@cpan.org>
381 * k81 is again client of k75 with different parameters.
383 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
384 about it all the time and nobody reacts accordingly; (2) lots of temp
385 files get created and not removed; culprit the new call to
386 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
387 caller has to remove the tempfile after use.
389 Need the drawing board.
391 PARTLY FIXED: retracting the idea to call
392 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
393 and unseed such that uptodate now means "we have mirrored this
394 recentfile's bunch of files" and seeded means "somebody believes the
395 index file for this rf needs be refreshed". This seems to work now.
397 Still collecting temp files that nobody cares to depollute.
399 2009-03-21 Andreas J. Koenig <andk@cpan.org>
401 * Bug?: should it be harder than it is atm to set the timestamp to the
402 future? YES, FIXED
404 * bug with native integers:
406 -
407 epoch: 997872011
408 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
409 type: new
410 -
411 epoch: 1195248431
412 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
413 type: new
415 Native integer broke when native math was turned off. FIXED
417 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
418 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
419 in this area.
421 Yes, records out of order:
423 447003 -
424 447004 epoch: 995885533
425 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
426 447006 type: new
427 447007 -
428 447008 epoch: 995890358
429 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
430 447010 type: new
431 447011 -
432 447012 epoch: 995892221
433 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
434 447014 type: new
436 FIXED with sanity check and later with the integer fix.
438 * Bug: want the index files in a .recent directory
440 * Bug: lots of dot files are not deleted in time
442 * possible test case: can a delete change the timestamp? This would
443 probably break the order of events.
445 2009-03-20 Andreas J. Koenig <andk@cpan.org>
447 * 1233701831.34486 what's so special about this number/string? It
448 belongs to
449 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
450 atm lives in Y,Q, and Z.
452 It is the first entry after id/--skip-locking which has timestamp
453 1234164228.11325 which represents a file that doesn't exist anymore.
455 * another bug:
457 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
458 _bigfloatcmp called with l[1237505213.21133]r[UNDEF]: but both must be defined at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
459 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
460 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
461 File::Rsync::Mirror::Recentfile::Done::covered('File::Rsync::Mirror::Recentfile::Done=HASH(0x8857fb4)', 1237505213.21133, 0.123456789) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile.pm line 2041
462 File::Rsync::Mirror::Recentfile::uptodate('File::Rsync::Mirror::Recentfile=HASH(0x8533a2c)') called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recent.pm line 536
463 File::Rsync::Mirror::Recent::rmirror('File::Rsync::Mirror::Recent=HASH(0x82ef3d0)', 'skip-deletes', 1) called at /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl line 27
464 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
465 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
466 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
467 File::Rsync::Mirror::Recentfile::Done::covered('File::Rsync::Mirror::Recentfile::Done=HASH(0x8857fb4)', 1237505213.21133, 0.123456789) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile.pm line 2041
468 File::Rsync::Mirror::Recentfile::uptodate('File::Rsync::Mirror::Recentfile=HASH(0x8533a2c)') called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recent.pm line 536
469 File::Rsync::Mirror::Recent::rmirror('File::Rsync::Mirror::Recent=HASH(0x82ef3d0)', 'skip-deletes', 1) called at /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl line 27
471 FIXED, it was the "--skip-locking" file where manual intervention was participating
473 * bug on the mirroring slave: when the dirtymark gets increased we
474 probably do not reset the done intervals. The mirrorer stays within
475 tight bounds where it tries to sync with upstream and never seems to
476 finish. In the debugging state file I see lots of identical intervals
477 that do not get collapsed. When I restart the mirrorer it dies with:
479 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
480 _bigfloatcmp called with l[1237400817.94363]r[UNDEF]: but both must be defined at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
481 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
482 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
483 File::Rsync::Mirror::Recentfile::Done::_register_one('File::Rsync::Mirror::Recentfile::Done=HASH(0x84c6af8)', 'HASH(0xb693f2dc)') called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 200
484 File::Rsync::Mirror::Recentfile::Done::register('File::Rsync::Mirror::Recentfile::Done=HASH(0x84c6af8)', 'ARRAY(0x8c54bfc)', 'ARRAY(0xb67e618c)') called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile.pm line 1044
485 File::Rsync::Mirror::Recentfile::_mirror_item('File::Rsync::Mirror::Recentfile=HASH(0x84abcb0)', 227, 'ARRAY(0x8c54bfc)', 33110, 'File::Rsync::Mirror::Recentfile::Done=HASH(0x84c6af8)', 'HASH(0x84abd8c)', 'ARRAY(0x839cb64)', 'HASH(0x839c95c)', 'HASH(0xb6a30f0c)', ...) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile.pm line 992
486 File::Rsync::Mirror::Recentfile::mirror('File::Rsync::Mirror::Recentfile=HASH(0x84abcb0)', 'piecemeal', 1, 'skip-deletes', 1) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recent.pm line 564
487 File::Rsync::Mirror::Recent::_rmirror_mirror('File::Rsync::Mirror::Recent=HASH(0x84ab6d4)', 7, 'HASH(0x8499488)') called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recent.pm line 532
488 File::Rsync::Mirror::Recent::rmirror('File::Rsync::Mirror::Recent=HASH(0x84ab6d4)', 'skip-deletes', 1) called at /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl line 27
490 and debugging stands at
492 At: before
493 Brfinterval: Z
494 Ci: 227
495 Dre+1:
496 epoch: 1237400802.5789
497 path: id/J/JO/JOHND/CHECKSUMS
498 type: new
499 Dre-0:
500 epoch: 1237400807.97514
501 path: id/M/MI/MIYAGAWA/CHECKSUMS
502 type: new
503 Dre-1:
504 epoch: 1237400817.94363
505 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
506 type: new
507 Eintervals:
508 -
509 - 900644040
510 - 900644040
511 - []
513 and it is reproducable.
515 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
516 pause has a fresh one.
518 FIXED, it was the third anded term in each of the ifs in the IV block in
519 _register_one: with that we make sure that we do not stamp on valuable
520 interval data.
522 2009-03-17 Andreas J. Koenig <andk@cpan.org>
524 * done: verified the existence of the floating point bug in bleadperl
525 and verified that switching from YAML::Syck to YAML::XS does not resolve
526 it.
528 BTW, the switch was doable with
530 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
532 and should be considered as a separate TODO
534 * todo: integrate a dirty update with two aggregate calls before
535 unlocking for frictionless dirtying DONE
537 * todo: start the second rsync daemon on pause DONE
539 * todo: move index files to .recent: this cannot simply be done by
540 setting filenameroot to .recent/RECENT. Other parts of the modules rely
541 on the fact that dirname(recentfile) is the root of the mirrored tree.
543 2009-03-16 Andreas J. Koenig <andk@cpan.org>
545 * What was the resolution of the mirror.pl delete hook bug? Do we call
546 the delete hook when pause removes a file from MUIR?
548 * Today on pause: Updating 2a13fba..29f284d and installing it for
549 /usr/local/perl-5.10.0{,-RC2}
551 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
552 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
554 2009-03-15 Andreas J. Koenig <andk@cpan.org>
556 * currently recent_events has the side effect of setting dirtymark
557 because it forces a read on the file. That should be transparent, so
558 that the dirtymark call always forces a cache-able(?) read.
560 * The bug below is -- after a lot of trying -- not reproducible on a
561 small script, only in the large test script. The closest to the output
562 below was:
564 #!perl
565 use strict;
566 use Devel::Peek;
567 my $x = "01237123229.8814";
568 my($l,$r);
569 for ($l,$r) {
570 $_ = "x"x34;
571 }
572 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
573 $r = int $r;
574 $l = "1237123231.22458";
575 $r = "1237123231.22458";
576 1 if $l/1.1;
577 Devel::Peek::Dump $l;
578 Devel::Peek::Dump $r;
579 Devel::Peek::Dump $x = $l <=> $r;
580 die "BROKE" if $x;
581 __END__
583 The checked in state at c404a85 fails the test with my
584 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
585 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
586 in the test script always consistent.
588 * Todo: write a test that inserts a second dirty file with an already
589 existing timestamp. DONE
591 * Bug in perl 5.10 on my 64bit box:
593 DB<98> Devel::Peek::Dump $l
594 SV = PVMG(0x19e0450) at 0x142a550
595 REFCNT = 2
596 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
597 IV = 0
598 NV = 1237123231.22458
599 PV = 0x194ce70 "1237123231.22458"\0
600 CUR = 16
601 LEN = 40
603 DB<99> Devel::Peek::Dump $r
604 SV = PVMG(0x19e0240) at 0x142a3e8
605 REFCNT = 2
606 FLAGS = (PADMY,POK,pPOK)
607 IV = 1237123229
608 NV = 1237123229.8814
609 PV = 0x19ff900 "1237123231.22458"\0
610 CUR = 16
611 LEN = 40
613 DB<100> Devel::Peek::Dump $l <=> $r
614 SV = IV(0x19ea6e8) at 0x19ea6f0
615 REFCNT = 1
616 FLAGS = (PADTMP,IOK,pIOK)
617 IV = -1
619 DB<101> Devel::Peek::Dump $l
620 SV = PVMG(0x19e0450) at 0x142a550
621 REFCNT = 2
622 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
623 IV = 0
624 NV = 1237123231.22458
625 PV = 0x194ce70 "1237123231.22458"\0
626 CUR = 16
627 LEN = 40
629 DB<102> Devel::Peek::Dump $r
630 SV = PVMG(0x19e0240) at 0x142a3e8
631 REFCNT = 2
632 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
633 IV = 1237123231
634 NV = 1237123231.22458
635 PV = 0x19ff900 "1237123231.22458"\0
636 CUR = 16
637 LEN = 40
639 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
641 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
643 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
645 * $obj->merge ($other) needs to learn about equal epoch which may happen
646 since dirty_epoch intruded.
648 * Wontfix anytime soon: I think we currently do not support mkdir. Only
649 files!
651 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
653 * Todo: continue working on update(...,$dirty_epoch). It must be
654 followed by a fast_aggregate! DONE
656 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
658 * maybe we need a closest_entry or fitting_interval or something like
659 that. We want to merge an event into the middle of some recentfile.
660 First we do not know which file, then we do not know where to lock,
661 where to enter the new item, when and where to correct the dirtymark.
663 So my thought is we should first find which file.
665 Another part of my brain answers: what would happen if we would enter
666 the new file into the smallest file just like an ordinary new event,
667 just as an old event?
669 (1) we would write a duplicate timestamp? No, this would be easy to
670 avoid
672 (2) we would make the file large quickly? Yes, but so what? We are
673 changing the dirtymark, so are willing to disturb the downstream hosts.
675 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
677 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
679 leak!
681 https://rt.cpan.org/Ticket/Display.html?id=41199
683 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
685 Strangest output being HAKANARDO who managed to upload
687 Here is a better oneliner that includes also the first line of each
688 finding:
690 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$F[-1]=~s/\s+\z//; push @{$Seen{$F[-1]}||=[]},$_; END {for my $k (keys %Seen){ delete $Seen{$k} if @{$Seen{$k}}==1; } print YAML::Syck::Dump(\%Seen)}'
692 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
694 * memory leak in the syncher? It currently weighs 100M.
696 Update 2008-11-02:
698 root 10705 1.0 4.9 80192 76596 pts/32 S+ Nov02 24:05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
701 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
703 * lookup by epoch and by path and use this ability on the pause to never
704 again register a file twice that doesn't need it. Let's call it
705 contains().
707 * after the dirtymark is done: fill up recentfiles with fake (historic)
708 entries; fill up with individual corrections; algorithm maybe to be done
709 with bigfloat so that we can always place something in the middle
710 between two entries. Before we must switch to bigfloat we could try to
711 use Data::Float::nextup to get the.
713 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
714 directly.
716 * dirtymark now lives in Recentfile, needs to be used in rmirror.
718 * find out why the downloader died after a couple of hours without a net
719 connection. Write a test that survives the not-existence of the other
720 end forever.
722 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
724 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
725 it makes less sense with tight coupling for secondary files. But asking
726 the server what the current epoch is might be cheaper on HTTP than on
727 rsync. (Needs to be evaluated)
729 * remove the 0.00 from the verbose overview in the Merged column in the
730 Z row. DONE
732 * write tests that expose the problems of the last few days: cascading
733 client/server roles, tight coupling for secondary RFs, deletes after
734 copies.
736 * Some day we might want to have policy options for the slave:
737 tight/loose/no coupling with upstream for secondary RFs. tight is what
738 we have now. loose would wait until a gap occurs that can be closed.
740 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
742 * revisit all $rfs->[$i+1] places if they now make sense still
744 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
746 * another bug is the fact that the mirror command deletes files before
747 it unhides the index file, thus confusing downstream slaves. We must not
748 delete before unhiding and must delete after unhiding. FIXED.
750 * new complication about the slave that is playing a server role.
751 Currently we mirror from newest to oldest with a hidden temporary file
752 as index. And when one file is finished, we unhide the index file.
753 Imagine the cascading server/slave is dead for a day. It then starts
754 mirroring again with the freshest thing and unhides the freshest index
755 file when it has worked through it. In that moment it exposes a time
756 hole. Because it now works on the second recentfile which is still
757 hidden.
759 We currently do nothing special to converge after such a drop out. At
760 least not intentionally and robustly and thought through.
762 The algorithm we use to seed the next file needs quite a lot of more
763 robustness than it currently has. Something to do with looking at the
764 merged element of the next rf and when it has dropped off, we seed
765 immediately. And if it ramains dropped off, we seed again, of course.
767 Nope, looking from smaller to larger RFS we look at the merged element
768 of this RF and at the minmax/max element of the next RF. If that
769 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
770 seeding it.
772 And we need a public accessor seed and unseed or seeded. But not the mix
773 of public and private stuff that then is used behind the back.
775 And then the secondary* stuff must go.
777 And we must understand what the impact is on the DONE system. Can it go
778 unnoticed that there was a hole? And could the DONE system have decided
779 the hole is covered? This should be testable with three directories where
780 the middle stops working for a while. Done->merge is suspicious, we must
781 stop it from merging non-conflatable neighbors due to broken continuity.
783 FIXED
785 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
787 * Slaven suggests to have the current epoch or the whole current
788 recentfile available from the HTTP server and take it away with
789 keepalive. This direction goes the granularity down to subseconds.
791 We might want to rewrite everything to factor out transport and allow
792 the whole thing to run via HTTP.
794 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
796 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
797 symlink?
799 * smoker on k81 fetching from k75 to verify cascading works. See
800 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
802 * maybe the loop should wait for CHECKSUMS file after every upload. And
803 CPAN.pm needs to deal with timestamps in the future.
805 * do not forget the dirtymark!
807 Text: have a new flag on recentfiles with the meaning: if this
808 changes, you're required to run a full rsync over all the files. The
809 reason why we set it would probably be: some foul happened. we injected
810 files in arbitrary places or didn't inject them although they changed.
811 The content of the flag? Timestamp? The relation between the
812 recentfiles would have to be inheritance from the principal, because any
813 out of band changes would soon later propagate to the next recentfile.
815 By upping the flag often one can easily ruin the slaves.
817 last out of band change? dirtymark?
819 Anyway, this implies that we read a potentially existing recentfile
820 before we write one.
822 And it implies that we have an eventloop that keeps us busy in 2-3
823 cycles, one for current stuff (tight loop) and one for the recentfiles
824 (cascade when principal has changed), one for the old stuff after a
825 dirtymark change.
827 And it implies that the out-of-band change in any of the recentfiles
828 must have a lock on the principal file and there is the place to set the
829 dirtymark.
831 * start a FAQ, especially quick start guide questions. Also to aid those
832 problematic areas where we have no good solution, like the "links"
833 option to rsync.
835 * wish feedback when we are slow.
837 * reduce mccabe
839 * Remove a few DEBUG statements.
841 * The multiple-rrr way of doing things needs a new option to rmirror,
842 like piecemeal or so. Not urgent because after the first pass through,
843 things run smoothely. It's only ugly during the first pass.
845 * I have the suspicion that the code is broken that decides if the
846 neighboring RF needs to be seeded. I fear when too much time has gone
847 between two calls (in our case more than one hour), it would not seed
848 the neighbor. Of course this will never be noticed, so we need a good
849 test for it.
851 * local/localroot confusion: I currently pass both options but one must
852 do.
854 * accounts for early birds on PAUSE rsync daemon.
856 * hardcoded 20 seconds
858 * who mirrors the index? DOING now.
860 * which CPAN mirrors offer rsync?
862 * visit all XXX, visit all _float places
864 * rename the pathdb stuff, it's too confusing. No idea how.
866 * rrr-inotify, backpan, rrr-register
868 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
870 * current bugs: the pathdb seems to get no reset, the seeding of the
871 secondaryttl stuff seems not to have an effect. Have helped myself with
872 a rand(10), need to fix this back. So not checked in. Does the rand
873 thing even help?
875 The rand thing helps. The secondaryttl stuff was in the wrong line,
876 fixed now.
878 The pathdb stuff was because I called either _pathdb or __pathdb on the
879 wrong object. FIXED now.
881 * It's not so beautiful if we never fetch the recentfiles that are not
882 the principal, even if this is correct behaviour. We really do not need
883 them after we have fetched the whole content.
885 OK, we want a switch for that: secondaryttl DONE
887 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
889 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
890 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
891 it lets 40 follow. FIXED
893 * See that the long running process really only updates the principal
894 file unless it has missed a timespan during which something happened. If
895 nothing happened, it must notice even when it misses the timespan. DONE
897 * we must throw away the pathdb when we have reached the end of Z. From
898 that moment we can have a very small pathdb because the only reason for
899 a pathdb is that we know to ignore old records in old files. We won't
900 need this pathdb again before the next full pass over the data is
901 necessary and then we will rebuild it as we go along. DONE
903 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
905 * I think, Done::register_one is doing wrong in that it does not
906 conflate neighboring pieces. The covered() method cannot do this because
907 it has no recent_events array at hand. But register_one has it and could
908 do it and for some reason misses to do it (sometimes).
910 This means that the three tests I just wrote can probably not survive
911 because they test with an already broken Done structure.
913 The art now is to detect how it happens, then to reproduce, then write a
914 test, then fix it.
916 So from the logfile this is what happens: we have a good interval with
917 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
918 top of it. Locally we now mirror F2 and open a new done interval for it.
919 Then we mirror F1 but this time with the timestamp T1b. And when we then
920 try to close the gap, we do not find T1 but instead something older. We
921 should gladly accept this older piece and this would fix this bug.
923 FIXED
925 * bug to fix: when the 1h file changes while rmirror is running, we do
926 correctly sync the new files but never switch to the 6h file but rather
927 stay in a rather quick loop that fetches the 1h file again and again.
929 Is it possible that we initialize a new object? Or does
930 get_remote_recentfile_as_tempfile overwrite something in myself?
932 Want a new option: _runstatusfile => $file which frequently dumps the
933 state of all recentfiles to a file.
935 FIXED
937 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
939 * Todo: now teach update to verify the timestamp is about to write
940 against the previous and use _increase_a_bit if it doesn't comply with
941 strict monotony. DONE
943 * The problem of rounding. So far perl's default precision was
944 sufficient. One day it won't be. FakeFloat has an easy job when it is
945 only reading and other machines have written correctly. But when we want
946 to write a floating point number that is a bit larger than the other
947 one, then we need our own idea of precision.
949 Slaven said: just append a "1". This might be going towards the end of
950 usability too quickly. I'd like something that actually uses the decimal
951 system. Well, appending a 1 also does this but...
953 E.g. we have 1.0. nextup on this architecture is starting with
954 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
955 taken the 1.0000000000000003 and the next user comes and the time tells
956 him 1.0 again. He has to beat my number without stepping over the
957 nextup. This is much less space than I had when I chose 1,2,3.
959 What is also irritating is that nextup is architecture dependent. The
960 128 bit guy must choose very long numbers to fit in between whereas the
961 other one with 16 bit uses larger steps. But then the algorithm is the
962 same for both, so that would be a nice thing.
964 I see two situation where we need this. One is when Time::HiRes returns
965 us a value that is <= the last entry in our recentfile. In this case
966 (let's call it the end-case) we must fill the region between that number
967 and the next higher native floating point number. The other is when we
968 inject an old file into an old recentfile (we would then also set a new
969 dirtymark). We find the integer value already taken and need a slightly
970 different one (let's call it the middle-case). The difference between
971 the two situations is that the next user will want to find something
972 higher than my number in the end-case and something lower than my number
973 in the middle case.
975 So I suggest we give the function both a value and an upper bound and it
976 calculates us a primitive middle. The upper bound in the middle-case is
977 the next integer. The upper bound on the end-case is the nextup floating
978 point number. But the latter poses another problem: if we have occupied
979 the middle m between x and nextup(x), then the nextup(m) will probably
980 not be the same as nextup(x) because some rounding will take place
981 before the nextup is calculated and when the rounding reaches the
982 nextup(x), we will end up at nextup(nextup(x)).
984 So we really need to consider the nextup and the nextdown from there and
985 then the middle and that's the number we may approach asymptotically.
986 Ugly. But DONE.
988 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
990 * consider deprecating the use of RECENT.recent as a symlink. It turns
991 out to need extra hoops with the rsync options and just isn't worth it.
992 Or maybe these extra hoops are needed anyway for the rest of the tree?
993 Nope, can't be the case because not all filesystems support symlinks.
995 But before doing the large step, I'll deprecate the call of
996 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
997 only introduced to resolve RECENT.recent and complicates the routine far
998 beyond what it deserves.
1000 DONE. Won't deprecate RECENT.recent, just moved its handling to the
1001 supervisor.
1003 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1005 * I think it's a bug that the rsync_option links must be set to true in
1006 order to support RECENT.recent and that nobody cares to set it
1007 automatically. Similar for ignore_link_stat_errors.
1009 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1011 * Todo: find all todos together and make a plan what is missing for a
1012 release.
1014 - verifytree or something like that. fsck maybe.
1016 - rersyncrecent, the script itself? What it do?
1018 - a way to only mirror the recentfiles without mirroring the whole
1019 remote system such that people can decide to mirror only partially see
1020 also 2008-08-30. .shadow-xxx directory? this also needed for a
1021 filesystem that is still incomplete and might need the mirrorfiles for
1022 lookup(?)
1024 - long living objects that mirror again and again. Inject something
1025 into ta, see how it goes over to tb.
1027 - how do we continue filling up the DONE system when we use an object
1028 for the second time? "fully covered" and "uptodate" or new terminology.
1030 - overview called on the wrong file should be understandable
1032 - the meta data field that must change when we fake something up so that
1033 the downstream people know they have to re-fetch everything.
1035 - how tolerant are we against missing files upstream? how do we keep
1036 track? there are legitimate cases where we did read upstream index right
1037 before a file got deleted there and then find that file as new and want
1038 it. There are other cases that are not self healing and must be tracked
1039 and bugreported.
1041 - how, exactly, do we have to deal with deletes? With rsync errors?
1043 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
1044 authors) failed: No such file or directory (2)
1046 The file above is a delete in 1h and a new in file 1M and the
1047 delete in the locally running rmirror did not get propagated to the 1M
1048 object. Bug. And the consequence is a standstill.
1050 It seems that a slave that works with a file below the principal needs
1051 to merge things all the way up to get rid of later deletes. Or keep
1052 track of all deletes and skip them later. So we need a trackdeletes.pm
1053 similar to the done.pm?
1055 see also 2008-08-20 about spurious deletes that really have no add
1056 counterpart and yet they are not wrong.
1058 - consider the effect when resyncing the recentfile takes longer than
1059 the time per loop. Then we never rsync any file. We need to diagnose
1060 that and force an increase of that loop time. But when we later are fast
1061 enough again because the net has recovered, then we need to switch back
1062 to original parameters. ERm, no, it's enough to keep syncing at least
1063 one file before refetching an index file.
1065 - remember to verify that no temp files are left lying around and the
1066 signal handler
1068 - status file for not long running jobs that want to track upstream with
1069 a, say, cronjob.
1071 - revisit all XXX _float areas and study Sub::Exporter DONE
1073 - persistent DB even though we just said we do not need it. Just for
1074 extended capabilities and time savings when, for example, upstream
1075 announces a reset and we get new recentfiles and could then limit
1076 ourselves to a subset of files (those that have a changed epoch) in a
1077 first pass and would only then do the loop to verify the rest. Or
1078 something.
1080 * Todo: aggregate files should know their feed and finding the principal
1081 should be done stepwise. (?)
1083 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
1084 left locks around. DONE
1086 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1088 * maybe extend the _overview so that it always says if and where the
1089 last file is in the next file and where the next event in the next rf
1090 would lie. No, don't like this anymore. REJECT
1092 * take the two new redundant tests out again, only the third must
1093 survive. DONE
1095 * Todo: add a sanity check if the merged structure is really pointing to
1096 a different rf and that this different rf is larger. DONE
1098 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1100 * now test, if they are overlapping. And test if there is a file in the
1101 next rf that would fit into this rf's interval.
1103 1h 1222324012.8474 1222322541.7963 0.4086
1104 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
1105 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
1106 1W 1222313218.3626 1221708477.5829 167.9835
1108 I suspect that somebody writes a merged timestamp without having merged
1109 and then somebody else relies on it.
1111 If aggregate is running, the intervals must not be extravagated, if it
1112 is not running, there must not be bounds, the total number of events in
1113 the system must be counted and must be controlled throughout the tests.
1114 That the test required the additional update was probably nonsense,
1115 because aggregate can cut pieces too. FIXED & DONE
1117 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1119 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1120 changed. FIXED
1122 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1124 * Most apparent bug at the moment is that the recentfiles are fetched
1125 too often. Only the principal should be fetched and if it has not
1126 changed, the others should not be refetched. ATM I must admit that I'm
1127 happy that we refetch more often than needed because I can more easily
1128 fix bugs while the thing is running.
1130 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1131 done but the done system does not know about it. The reason for the
1132 failure is not known and we never reach the status uptodate because of
1133 this. We must get over it.
1135 Later it turns out that the origin server had a bug somewhere.
1136 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1137 array of the recentfile one position above. The bug was my own.
1139 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1141 * There is the race condition where the server does a delete and the
1142 slave does not yet know and then tries to download it because he sees
1143 the new. So for this time window we must be more tolerant against
1144 failure. If we cannot download a file, we should just skip it and should
1145 not retry immediately. The whole system should discover the lost thing
1146 later. Keeping track with the DONE system should really be a no brainer.
1148 But there is something more: the whole filesystem is a database and the
1149 recentfiles are one possible representation of it. It's a pretty useful
1150 representation I think that's why I have implemented something around
1151 it. But for strictly local operation it has little value. For local
1152 operation we would much rather have a database. So we would enter every
1153 recentfile reading and every rsync operation and for every file the last
1154 state change and what it leads to. Then we would always ignore older
1155 records without the efforts involved with recentfiles.
1157 The database would have: path,recentepoch,rsyncedon,deletedon
1159 Oh well, not yet clear where this leads to.
1161 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1163 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1164 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1166 * statusfile, maybe only on demand, alone to have a sharp debugging
1167 tool. It is locked and all recentfiles dump themselves into it and we
1168 can build a viewer that lets us know where we stand and what's inside.
1170 * remember: only the principal recentfile needs expiration, all others
1171 shall be expired by principal if it discovers that something has move
1172 upstream.
1174 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1176 * Always check if we stringify to a higher value than in the entry
1177 before. DONE
1179 * And in covered make an additional check if we would be able to see a
1180 numerical difference between the two numbers and if we can't then switch
1181 to a different, more expensive algorithm. Do not want to be caught by
1182 floating surprises. DONE
1184 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1186 * caching has several aspects here: we can cache the interval of the
1187 recentfile which only will change when the mtime of the file changes. We
1188 must re-mirror the recentfile when its ttl has expired. Does have_read
1189 tell you anything? It counts nothing at all. Only the mtime is
1190 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1191 thing the have_mirrored because it is unrelated to the mtime.
1193 * Robustness of floating point calculations! I always thought that the
1194 string calculated by the origin server for the floating representation
1195 of the epoch time is just a string. When we convert it to a number and
1196 later back to a string, the other computer might come to a different
1197 conclusion. This must not happen, we want to preserve it under any
1198 circumstances. I will have to write tests with overlong sequences that
1199 get lost in arithmetic and must see if all still works well. DONE
1201 But one fragile point remains: if one host considers a>b and the other
1202 one considers them == but no eq. To prevent this, we must probably do
1203 some extra homework. DONE
1205 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1207 * the concept of tracking DONE needs an object per recentfile that has
1208 something like these methods:
1210 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1212 covered() register() covered()
1214 The unclear thing is how we translate points in time into intervals. We
1215 could pass a reference to the current recent_events array when running
1216 we_have(xxx) and let the DONE object iterate over it such that it only
1217 has to store a list of intervals that can melt into each other. Ah, even
1218 passing the list together with a list of indexes seems feasiable.
1220 Or maybe ask for the inverted list?
1222 Whenever the complete array is covered by the interval we say we are
1223 fully covered and if the recentfile is not expired, we are uptodate.
1225 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1227 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1229 * need a way to "return" the next entry after the end of a list. When
1230 the caller says "before" or "after" we would like to know if he could
1231 cover that interval/threshold or not because this influences the effect
1232 of a newer timestamp of that recentfile. DONE with $opt{info}.
1234 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1236 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1238 loop implies tixu (?). Nope, something like --statefile decides. Per
1239 default we do ...?
1241 T test, I init, X including removals, U nodirtymark
1243 So we have no concept of dirtymarks, we only trust that since we are
1244 running we have observed everything steadily. But people will not let
1245 this program run forever so we must consider both startup penalty and
1246 book keeping for later runs. We keep this for later. For now we write a
1247 long running mirror that merges several intervals.
1249 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1251 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1252 the intervals!
1254 * rersyncrecent, the script: default to one week. The name of the switch
1255 is --after. Other switches? --loop!
1257 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1259 * need a switch --skip-deletes (?)
1261 * need a switch --enduser that tells us that the whole tempfile
1262 discipline is not needed when there is no downstream user. (?)
1264 Without this switch we cannot have a reasonable recent.pl that just
1265 displays the recent additions. Either we accept to download everything.
1266 Or we download temporary files without the typical rsync protocol
1267 advantages.
1269 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1270 File::Temp::tempdir, this might be a win.
1272 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1274 * apropos missing: we have no push, we never know the downstream
1275 servers. People who know their downstream hosts and want to ascertain
1276 something will want additional methods we have never thought about, like
1277 update or delete a certain file.
1279 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1281 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1282 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1283 difference.
1285 * rsync options: it's a bit of a pain that we usually need several rsync
1286 options, like compress, links, times, checksum and that there is no
1287 reasonable default except the original rsync default. I think wee can
1288 safely assume that the rsync options are shared between all recentfile
1289 instances within one recent tree.
1291 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1293 * deletes: if a delete follows an add quickly enough it may happen that
1294 a downstream mirror did not see the add at all! It seems this needs to
1295 be mentioned somewhere. The point here is that even if the downstream is
1296 never missing the principal timeframe it may encounter a "delete" that
1297 has no complimentary "add" anywhere.
1299 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1301 * I suspect the treat of metadata is incorrect during read or something.
1302 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1303 contained more than 6 hours worth of data. At 06:08 we merged into the
1304 1d file. We need to take snapshots of the 6h file over the course of an
1305 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1306 enough.
1308 Much worse: watching the 1h file: right at the moment (at 06:35) it
1309 covers 1218867584-1219120397 which is 70 hours.
1311 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1312 08:19:44 2008, that is when I checked out last time, so it seems to be
1313 aggregating and never truncating?
1315 No, correct is: it is never truncating; but wrong is: it is aggregating.
1316 It does receive a lot of events from time to time from a larger file.
1317 Somehow a large file gets merged into the small one and because the
1318 "meta/merged" attribute is missing, nobody is paying attention. I
1319 believe that I can fix this by making sure that metadata are honoured
1320 during read. DONE and test adjusted.
1322 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1324 * grand renaming plan
1326 remotebase => remoteroot to fit well with localroot DONE
1327 local_path() => localroot seems to me should already work DONE
1328 recentfile_basename => rfilename no need to stress it has no slash DONE
1330 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1331 kern? It would anyway need a deprecation cycle because it is an
1332 important constructor.
1334 * I like the portability that Data::Serializer brings us but the price
1335 is that some day we might find out that it is slowing us a bit. We'll
1336 see.
1338 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1340 * should we not enter the interval of the principal (or the interval of
1341 the merging file?) in every aggregated/merged file?
1343 * we should aim at a first release and give up on thinking about
1344 sanitizing stuff and zloop. Let's just admit that a full traditional
1345 rsync is the only available sanitizer ATM. Otherwise it's complicated
1346 stuff: sanitizing on the origin server, sanitizing on the slaves,
1347 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1348 get the basics out before this becomes a major cause for mess.
1350 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1352 * On OSes not supporting symlinks we expect that RECENT.recent contains
1353 the contents of the principal recentfile. Actually this is identical on
1354 systems supporting symlinks. Simple, what follows from that is that we
1355 need to keep the serializer in the metadata because we cannot read it
1356 from the filename, doesn't it? Of course not. It's a chicken and egg
1357 problem. This leaves us with the problem to actually parse the
1358 serialized data to find out in which format it is. So who can do the 4
1359 or 5 magics we wanted to support? File::LibMagic?
1361 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1363 * remotebase and recentfile_basename are ugly names. Now that we need a
1364 word for the shortest/principal/driving recentfile too we should do
1365 something about it.
1367 localroot is good. rfile is good. local_path() is bad, local_path($path)
1368 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1369 already deprecated.
1371 Up to now remotebase was the string that described the remote root
1372 directory in rsync notation, like pause.perl.org::authors. And
1373 recentfile_basename was "RECENT-1h.yaml".
1375 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1377 * The test that was added in today's checkin is a good start for a test
1378 of rmirror. We should have more methods in Recent.pm: verify,
1379 addmissingfiles. We should verify the current tree, then rmirror it and
1380 then verifytree the copy. We could then add some arbitrary file and let
1381 it be discovered by addmissingfiles, then rmirror again and then
1382 verifytree the copy again.
1384 Then we could start stealing from csync2 sqlite database [no port to
1385 OSX!] and fill a local DB. And methods to compare the database with the
1386 recentfiles. Our strength is that in principle we could maintain state
1387 with a single float. We have synced up to 1234567890.123456. If the Z
1388 file does not add new files all we have to do is mirror the new ones and
1389 delete the goners.
1391 This makes it clear that we should extend current protocol and declare
1392 that we cheat when we add files too late, just to help the other end
1393 keeping track. Ah yes, that's what was meant when zloop was mentioned
1394 earlier.
1396 Maybe need to revisit File::Mirror to help me with this task.
1398 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1400 * There must be an allow-me-to-truncate flag in every recentfile.
1401 Without it one could construct a sequence of updates winning the locking
1402 battle against the aggregator. Only if an aggregator has managed to
1403 merge data over to the next level, truncating can be allowed. DONE with
1404 accessor merged.
1406 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1408 * We should probably guarantee that no duplicates enter the aggregator
1409 array.
1411 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1413 * To get merge operation faster would need a good benchmark test. What
1414 02 spits out isn't reliable enough and is dominated by many other
1415 things. Between
1417 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1418 Author: Andreas J. Koenig <k@k75.(none)>
1419 Date: Sat Aug 2 07:58:04 2008 +0200
1421 and
1423 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1424 Author: Andreas J. Koenig <k@k75.(none)>
1425 Date: Sat Aug 2 11:40:29 2008 +0200
1427 there was a lot of trying but the effect is hardly measurable with
1428 current tests.
1430 * overhead of connecting seems high. When setting
1431 max_files_per_connection to 1 we see that.
1433 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1435 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1437 25138.2215600014/3600 = 6.98283932222261
1439 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1440 be a bug in the aggregator. FIXED
1442 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1444 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1445 we have it, can we mirror it, mirror it!
1447 I fear this needs a new class which might be called
1448 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1449 and treat them as an entity. I realize that a single recentfile may be
1450 sufficient for certain tasks and that it is handy for the low level
1451 programmer but it is not nice to use. If there is a delete in the 1h
1452 file then the 6h file still contains it. Seekers of the best information
1453 need to combine at least some of the recentfiles most of the time.
1455 There is the place for the Z loop!
1457 But the combination is something to collect in a database, isn't it. Did
1458 csync2 just harrumph?
1460 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1462 * it just occurred to me that hosts in the same mirroring pool could
1463 help out each other even without rewriting the recentfile. Just fetch
1464 the stuff to mirror from several places, bingo. But that's something
1465 that should rather live in a separate package or in rsync directly.
1467 * cronjobs are unsuited because with ntp they would all come at the full
1468 minute and disturb each other. Besides that I'd hate to have a backbone
1469 with more than a few seconds latency.
1471 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1473 * a second rsync server with access control for PAUSE. Port? 873 is the
1474 standard port, let's take 8873.
1476 * if there were a filesystem based on this, it would have a slow access
1477 to inexistent files. It would probably provide wrong readdir (only based
1478 on current content) or also a slow one (based on a recentfile written
1479 after the call). But it would provide fast access to existing files. Or
1480 one would deliberately allow slightly blurred answers based on some
1481 sqlite reflection of the recentfiles.
1483 * todo: write a variant of mirror() that combines two or more
1484 recentfiles and treats them like one
1486 * todo: signal handler to remove the tempfile
1488 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1490 * now that we have the symlink I forgot how it should be used in
1491 practice.
1493 * the z loop: add missing files to Z file. Just append them (instead of
1494 prepending). So one guy prepends something from the Y file from time to
1495 time and another guy appends something rather frequently. Collecting
1496 pond. When Y merges into Z, things get epoch and the collecting pond
1497 gets smaller. What exactly are "missing files"?
1499 take note of current epoch of the alpha file, let's call it the
1500 recent-ts
1502 find all files on disk
1504 remove all files registered in the recentworld up to recent-ts
1506 remove all files that have been deleted after recent-ts according to
1507 recentworld
1509 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1511 * rersyncrecent might be a cronjob with a (locked) state file which
1512 contains things like after and maybe last z sync or such?
1514 rrr-mirror might be an alternative name but how would we justify the
1515 three Rs when there is no Re-Rsync-Recent?
1517 With the --loop parameter it is an endless loop, without it is no loop.
1518 At least this is simple.
1520 * todo: new accssor z-interval specifies how often the Z file is updated
1521 against the filesystem. We probably want no epoch stamp on these
1522 entries. And we want to be able to filter the entries (e.g. no
1523 by-modules and by-category tree)
1525 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1527 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1528 into Z. We do not need the other files filled up with filesystem
1529 contents though.
1531 * need interface to query for a file in order to NOT call update on
1532 PAUSE a second time within a short time.
1534 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1536 * recommended update interval? Makes no sense, is different for
1537 different users.
1539 * Moosify
1541 Local Variables:
1542 mode: change-log
1543 change-log-default-name: "Todo"
1544 tab-width: 2
1545 left-margin: 2
1546 End: