| blob | 6bd5ef00addc00d8bee8b55fe14ba826067fab21 |
1 2011-02-20 Andreas J. Koenig <andk@cpan.org>
3 * start consistently considering the RECENT files themselves not part of
4 the set.
6 2011-02-19 Andreas J. Koenig <andk@cpan.org>
8 * Found the term anonymous one way file system
10 * No support for empty directories
12 2011-02-17 Andreas J. Koenig <andk@cpan.org>
14 * .recent/ or .rrr/ for metadata
16 * ALERT rrr: we have a piece of code somewhere in the mirror()
17 subroutine that reads the raw YAML file with normal open() and cuts it
18 off for efficiency, then feeds it to yaml::load in order to get some
19 metadata. Reading the thing with the whole tail of the recent array
20 kills the whole idea. The whole thing will not work with JSON.
22 So we must invent a protocol 2 that works with one additional index
23 files, say "A", only containing meta.
25 2011-02-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
27 * I think interval guaranteed with the $last_aggregate_call
28 variable (both in fsck and server) need to be fixed with a $recc
29 variable to something like 60 seconds or even less.
31 * minor bug: fsck just added the lockfile to the index which should be
32 considered bookkeeping.
34 * speed: a large directory remove that is done by the kernel in 1
35 seconds triggers minutes of bookkeeping (07:56:42 - 08:01:10 for ~3000
36 files, all due to Schlemiehl again. Want to do some collecting of
37 pending ops before actually writing to the RECENT files. Lock!
39 Repeating the timing with 2600 files and it took 21:28:08 - 21:35:27.
41 Now having rewritten the loop to use batch_update(): 4383 file removed
42 21:43:37 - 21:44:12. From 7:30 to 0:35 while doing 60% more work that's
43 a joy.
45 * lockdirectory expiration? server died and blocked fsck for so long.
47 * At the moment rrr-init is not needed anymore, rrr-server can be
48 started and followed by rrr-fsck. Still stupid but not the main
49 showstopper.
51 2011-02-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
53 * IN_Q_OVERFLOW needs to trigger an fsck, likewise the entry into the
54 server.
56 2011-01-30 Andreas J. Koenig <andk@cpan.org>
58 * rrr-init and rrr-server need to be used such:
60 init
61 server
62 init again
64 Stupid but no showstopper ATM.
66 2010-10-26 Andreas J. Koenig <andk@cpan.org>
68 * forking slave is done, needs more testing.
70 2010-07-06 Andreas J. Koenig <andk@cpan.org>
72 * The two next most important goals are to get a useable rrr-server and
73 a forking slave implementation that uses virtually no memory after the
74 first pass through. Does it need the status file?
76 * rrr-init now being fast we can now remove the alpha alerts. Done.
78 2010-06-16 Andreas J. Koenig <andk@cpan.org>
80 * rrr-init seems to be sssslllloooowwww. At 3:49 I started it for the
81 video tree. After twelve minutes the file has reached a size of 500k. It
82 is locking and rewriting the YAML after every file. Wasn't there an
83 unlocked variant? For "init" one would not expect locked correctness and
84 a rewrite after every item. Given that this tends to become slower with
85 every file it is really unacceptable.
87 Reminds me of rrr-dirtyupdate (the former name of rrr-update). It seems
88 the code was not dirty enough and that was the reason why it was so
89 slow. And rrr-update does not have a --dirty switch, so it looks like we
90 have no way to force dirty and fast operation yet but we need it for
91 init.
93 Irritating the bootstrapping race: the order {(1) fill initial
94 array, (2) start server} leaves the need for an fsck to cover the time
95 between 1 and 2. Given that regular fsck is needed in any case the price
96 is not high but still doesn't give a warm feeling. Maybe starting the
97 server will by itself do an fsck?
99 2010-06-15 Andreas J. Koenig <andk@cpan.org>
101 * remove the wild alpha alerts and say, it is alpha but parts of it are
102 already working very well or so.
104 2009-12-10 Andreas J. Koenig <andk@cpan.org>
106 * Prominent request for statusfile to safe memory consumption. I'd like
107 to work backwards: get a verification program that verifies how well in
108 sync two nodes are without using frmr itself. Pretend that the forking
109 agent already works and can be interrupted with ^C. Then let it die at
110 random and restart at random and see the memory consumption and the
111 failure rate over time. Compare it with the curve we have now.
113 Of course we also invented the status file as a sharp debugging tool.
114 I'd like to see it in action. See also: F:R:M:Recent::thaw().
116 2009-06-22 Andreas J. Koenig <andk@cpan.org>
118 * Todo: rrr-init to generate some index files so that rrr-server can be
119 started.
121 2009-05-02 Andreas J. Koenig <andk@cpan.org>
123 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
124 so we can use it for normal update with a single file. And add a cronjob
125 that runs rrr-fsck so we get a reminder when something is not working
126 as it should.
128 2009-04-27 Andreas J. Koenig <andk@cpan.org>
130 * Todo: I still see more link_stat errors than expected. I come back to
131 the former suspicion that we do not do necessary bookkeeping on deleted
132 files so that when we reach the "new" event on a meanwhile deleted file
133 we try to mirror it although we could know better. It's just noise, no
134 harm but noise is irritating.
136 * Todo: more experimenting with runstatusfile to get it official
138 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
139 parse the error.
141 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
143 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
145 And that ignore_link_stat_error is propagated in sparse_clone.
147 DONE
149 For the record: the bug was that ignore_link_stat_error was lost during
150 sparse_clone. And then the second bug was that it did not default to
151 true. Both are now fixed.
153 2009-04-25 Andreas J. Koenig <andk@cpan.org>
155 * Todo: fill rrr program with life
157 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
159 * Minor bug: I see complaints about files not existing like
160 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
161 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
163 In the RECENT files these are delete events but of course also new
164 events in older recent files.
166 I think we do not properly remember deletes when we skip-deletes and so
167 do not filter them out when we later see the "new" event.
169 Yes, but that what ignore_link_stat_error is for, we know about this
170 race condition. We must make ignore_link_stat_error default to true.
172 DONE
174 * bug in Done.pm failing to merge two fields into one when a third is
175 present. FIXED and accompanied with new test.
177 2009-04-24 Andreas J. Koenig <andk@cpan.org>
179 * Todo: keep the index clean and run some kind of nondestructive fsck 4
180 times a day.
182 * who is our backbone?
184 http://cpan.cpantesters.org/authors/02STAMP # barbie
185 http://cpan.solfo.com/authors/02STAMP # abh
186 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
188 2009-04-17 Andreas J. Koenig <andk@cpan.org>
190 * the changelog helper I included in the Makefile release memo does not
191 show tags and is much less useful than I thought. gitk is probably much
192 more convenient.
194 2009-04-16 Andreas J. Koenig <andk@cpan.org>
196 * ABH writes:
198 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
199 #### Ydr_.yaml ... DONE
200 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
201 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
202 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
203 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
204 #### gone already at cpan-pause.pl line 0
206 Running without skip-deletes now but cannot reproduce.
208 2009-04-15 Andreas J. Koenig <andk@cpan.org>
210 * consider state directory such we can restart after a ^C or come again
211 from a cronjob.
213 Reuse _runstatusfile: write atomically, see how we can use the dump to
214 start running again. Do not leak and remove globs before writing. Do not
215 only dump one rf, dump the whole complex.
217 * consider the effect of ^C. Is it safe?
219 * want to use Perl::Version to up the versions on every release. Only if
220 things have changed, of course. Does this actually work?
222 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
223 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
225 Must do the bookkeeping myself when this should happen, maybe based on git?
227 I think I prefer the same version number for all pm files. But one also
228 needs a routine for setting them. Ahh:
230 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
232 NEEDSMOREWORK
234 * why indexer wrong?
236 cpan[2]> m /Mirror::Recent/
237 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
238 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
239 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
240 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
241 4 items found
243 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
245 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
247 I just tried MM 6.5102 and it again produces only a provides for Recent,
248 not for the other modules.
250 My own fault. FIXED in the Makefile.PL.
252 2009-04-13 Andreas J. Koenig <andk@cpan.org>
254 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
256 * is it true (as stated at
257 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
258 that rsync can lead to errors when upstream changes before client sync
259 has completed? Can we see the error when we run this on upstream:
261 % perl -e '
262 use Time::HiRes qw(time);
263 while (){
264 open my $fh, ">", "changingfile.txt.new" or die;
265 print $fh (time."\n") x 1000000;
266 close $fh;
267 rename "changingfile.txt.new", "changingfile.txt" or die;
268 }
269 '
271 And on the receiving end:
273 % while true; do
274 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
275 sleep 1;
276 done
278 I cannot get it to fail. Apparently it is sufficient when upstream
279 always writes atomically (which of course is mandatory)?
281 SENT to mailing list Apr 28.
283 * from ABH:
285 > https://fedorahosted.org/InstantMirror/
286 > https://www.redhat.com/mailman/listinfo/instantmirror-list
287 > irc.freenode.net channel #instantmirror
289 A cool name for a project. Inspires me to write a few sentences about
290 bittorrent's role in the grand picture.
292 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
293 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
295 2009-04-12 Andreas J. Koenig <andk@cpan.org>
297 * Interesting last minute bug during real download testing: the output
298 isn't as pretty anymore, it seems that more work is being done than
299 needed.
301 Introducing a highlevel _alluptodate method helped a lot in debugging
302 this.
304 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
305 different on 1h file than on the other files. This lead to frequent
306 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
307 issue.
309 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
311 * Todo: try if we can be faster with a native float package. I'm really
312 glad to have a machine exposing perl floating point bugs, it forced me
313 to retract this item from the changes file: "several speedups in the
314 fakebigfloat code" before I get cpan testers fails.
316 2009-04-11 Andreas J. Koenig <andk@cpan.org>
318 * Bug: when the dirtymark changes upstream, then the downstream server
319 notices it immediately and mirrors the "1h" file but then leaves the
320 rmirror loop and continues with "6h" on the next round through. This
321 behaviour goes on until it reaches the Z file. Because we eat the Z file
322 piecemeal we leave the loop after a while but after that everybody seems
323 to have forgotten that there is some work left to be done. Bad, bad,
324 bad.
326 ETOOCONFUSING better name?
328 Recentfile::get_remote_recentfile_as_tempfile OK
329 Recentfile::resolve_recentfilename split_rfilename DONE
331 Recent::_principal_recentfile_object _principal_recentfile_fromremote
332 Recent::principal_recentfile OK
333 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
335 some void, some not, some not void but called in void context.
337 2009-04-10 Andreas J. Koenig <andk@cpan.org>
339 * Bug? When the dirtymark changes then the second tier hosts quickly
340 reset their DONE state and restart mirroring. But then they mirror
341 potentially outdated index files with inconsistent dirtymark because the
342 first tier box may be a bit behind on the large recentfiles.
344 This means we need a brake when iterating over the recentfiles that
345 refuses to use a recentfile with the wrong dirtymark.
347 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
349 * install the bin/* files and talk about rrr-overview somewhere.
351 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
352 created outside the target tree.
354 * Barbie asks for better logging capabilities
356 2009-03-29 Andreas J. Koenig <andk@cpan.org>
358 * known bug: cannot configure the Recentfile objects to keep delete
359 events. How to call the accessor for that? keep_delete_objects_forever?
361 Sounds acceptable. FIXED (but untested).
363 * known bug: keeps temporary index files lying around.
365 2009-03-24 Andreas J. Koenig <andk@cpan.org>
367 * Some equivalent for
369 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
371 ? DONE
373 2009-03-22 Andreas J. Koenig <andk@cpan.org>
375 * k81 is again client of k75 with different parameters.
377 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
378 about it all the time and nobody reacts accordingly; (2) lots of temp
379 files get created and not removed; culprit the new call to
380 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
381 caller has to remove the tempfile after use.
383 Need the drawing board.
385 PARTLY FIXED: retracting the idea to call
386 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
387 and unseed such that uptodate now means "we have mirrored this
388 recentfile's bunch of files" and seeded means "somebody believes the
389 index file for this rf needs be refreshed". This seems to work now.
391 Still collecting temp files that nobody cares to depollute.
393 2009-03-21 Andreas J. Koenig <andk@cpan.org>
395 * Bug?: should it be harder than it is atm to set the timestamp to the
396 future? YES, FIXED
398 * bug with native integers:
400 -
401 epoch: 997872011
402 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
403 type: new
404 -
405 epoch: 1195248431
406 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
407 type: new
409 Native integer broke when native math was turned off. FIXED
411 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
412 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
413 in this area.
415 Yes, records out of order:
417 447003 -
418 447004 epoch: 995885533
419 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
420 447006 type: new
421 447007 -
422 447008 epoch: 995890358
423 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
424 447010 type: new
425 447011 -
426 447012 epoch: 995892221
427 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
428 447014 type: new
430 FIXED with sanity check and later with the integer fix.
432 * Bug: want the index files in a .recent directory
434 * Bug: lots of dot files are not deleted in time
436 * possible test case: can a delete change the timestamp? This would
437 probably break the order of events.
439 2009-03-20 Andreas J. Koenig <andk@cpan.org>
441 * 1233701831.34486 what's so special about this number/string? It
442 belongs to
443 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
444 atm lives in Y,Q, and Z.
446 It is the first entry after id/--skip-locking which has timestamp
447 1234164228.11325 which represents a file that doesn't exist anymore.
449 * another bug:
451 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
452 _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
453 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
454 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
455 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
456 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
457 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
458 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
465 FIXED, it was the "--skip-locking" file where manual intervention was participating
467 * bug on the mirroring slave: when the dirtymark gets increased we
468 probably do not reset the done intervals. The mirrorer stays within
469 tight bounds where it tries to sync with upstream and never seems to
470 finish. In the debugging state file I see lots of identical intervals
471 that do not get collapsed. When I restart the mirrorer it dies with:
473 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
474 _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
475 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
476 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
477 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
478 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
479 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
480 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
481 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
482 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
484 and debugging stands at
486 At: before
487 Brfinterval: Z
488 Ci: 227
489 Dre+1:
490 epoch: 1237400802.5789
491 path: id/J/JO/JOHND/CHECKSUMS
492 type: new
493 Dre-0:
494 epoch: 1237400807.97514
495 path: id/M/MI/MIYAGAWA/CHECKSUMS
496 type: new
497 Dre-1:
498 epoch: 1237400817.94363
499 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
500 type: new
501 Eintervals:
502 -
503 - 900644040
504 - 900644040
505 - []
507 and it is reproducable.
509 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
510 pause has a fresh one.
512 FIXED, it was the third anded term in each of the ifs in the IV block in
513 _register_one: with that we make sure that we do not stamp on valuable
514 interval data.
516 2009-03-17 Andreas J. Koenig <andk@cpan.org>
518 * done: verified the existence of the floating point bug in bleadperl
519 and verified that switching from YAML::Syck to YAML::XS does not resolve
520 it.
522 BTW, the switch was doable with
524 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
526 and should be considered as a separate TODO
528 * todo: integrate a dirty update with two aggregate calls before
529 unlocking for frictionless dirtying DONE
531 * todo: start the second rsync daemon on pause DONE
533 * todo: move index files to .recent: this cannot simply be done by
534 setting filenameroot to .recent/RECENT. Other parts of the modules rely
535 on the fact that dirname(recentfile) is the root of the mirrored tree.
537 2009-03-16 Andreas J. Koenig <andk@cpan.org>
539 * What was the resolution of the mirror.pl delete hook bug? Do we call
540 the delete hook when pause removes a file from MUIR?
542 * Today on pause: Updating 2a13fba..29f284d and installing it for
543 /usr/local/perl-5.10.0{,-RC2}
545 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
546 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
548 2009-03-15 Andreas J. Koenig <andk@cpan.org>
550 * currently recent_events has the side effect of setting dirtymark
551 because it forces a read on the file. That should be transparent, so
552 that the dirtymark call always forces a cache-able(?) read.
554 * The bug below is -- after a lot of trying -- not reproducible on a
555 small script, only in the large test script. The closest to the output
556 below was:
558 #!perl
559 use strict;
560 use Devel::Peek;
561 my $x = "01237123229.8814";
562 my($l,$r);
563 for ($l,$r) {
564 $_ = "x"x34;
565 }
566 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
567 $r = int $r;
568 $l = "1237123231.22458";
569 $r = "1237123231.22458";
570 1 if $l/1.1;
571 Devel::Peek::Dump $l;
572 Devel::Peek::Dump $r;
573 Devel::Peek::Dump $x = $l <=> $r;
574 die "BROKE" if $x;
575 __END__
577 The checked in state at c404a85 fails the test with my
578 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
579 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
580 in the test script always consistent.
582 * Todo: write a test that inserts a second dirty file with an already
583 existing timestamp. DONE
585 * Bug in perl 5.10 on my 64bit box:
587 DB<98> Devel::Peek::Dump $l
588 SV = PVMG(0x19e0450) at 0x142a550
589 REFCNT = 2
590 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
591 IV = 0
592 NV = 1237123231.22458
593 PV = 0x194ce70 "1237123231.22458"\0
594 CUR = 16
595 LEN = 40
597 DB<99> Devel::Peek::Dump $r
598 SV = PVMG(0x19e0240) at 0x142a3e8
599 REFCNT = 2
600 FLAGS = (PADMY,POK,pPOK)
601 IV = 1237123229
602 NV = 1237123229.8814
603 PV = 0x19ff900 "1237123231.22458"\0
604 CUR = 16
605 LEN = 40
607 DB<100> Devel::Peek::Dump $l <=> $r
608 SV = IV(0x19ea6e8) at 0x19ea6f0
609 REFCNT = 1
610 FLAGS = (PADTMP,IOK,pIOK)
611 IV = -1
613 DB<101> Devel::Peek::Dump $l
614 SV = PVMG(0x19e0450) at 0x142a550
615 REFCNT = 2
616 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
617 IV = 0
618 NV = 1237123231.22458
619 PV = 0x194ce70 "1237123231.22458"\0
620 CUR = 16
621 LEN = 40
623 DB<102> Devel::Peek::Dump $r
624 SV = PVMG(0x19e0240) at 0x142a3e8
625 REFCNT = 2
626 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
627 IV = 1237123231
628 NV = 1237123231.22458
629 PV = 0x19ff900 "1237123231.22458"\0
630 CUR = 16
631 LEN = 40
633 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
635 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
637 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
639 * $obj->merge ($other) needs to learn about equal epoch which may happen
640 since dirty_epoch intruded.
642 * Wontfix anytime soon: I think we currently do not support mkdir. Only
643 files!
645 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
647 * Todo: continue working on update(...,$dirty_epoch). It must be
648 followed by a fast_aggregate! DONE
650 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
652 * maybe we need a closest_entry or fitting_interval or something like
653 that. We want to merge an event into the middle of some recentfile.
654 First we do not know which file, then we do not know where to lock,
655 where to enter the new item, when and where to correct the dirtymark.
657 So my thought is we should first find which file.
659 Another part of my brain answers: what would happen if we would enter
660 the new file into the smallest file just like an ordinary new event,
661 just as an old event?
663 (1) we would write a duplicate timestamp? No, this would be easy to
664 avoid
666 (2) we would make the file large quickly? Yes, but so what? We are
667 changing the dirtymark, so are willing to disturb the downstream hosts.
669 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
671 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
673 leak!
675 https://rt.cpan.org/Ticket/Display.html?id=41199
677 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
679 Strangest output being HAKANARDO who managed to upload
681 Here is a better oneliner that includes also the first line of each
682 finding:
684 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
686 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
688 * memory leak in the syncher? It currently weighs 100M.
690 Update 2008-11-02:
692 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
695 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
697 * lookup by epoch and by path and use this ability on the pause to never
698 again register a file twice that doesn't need it. Let's call it
699 contains().
701 * after the dirtymark is done: fill up recentfiles with fake (historic)
702 entries; fill up with individual corrections; algorithm maybe to be done
703 with bigfloat so that we can always place something in the middle
704 between two entries. Before we must switch to bigfloat we could try to
705 use Data::Float::nextup to get the.
707 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
708 directly.
710 * dirtymark now lives in Recentfile, needs to be used in rmirror.
712 * find out why the downloader died after a couple of hours without a net
713 connection. Write a test that survives the not-existence of the other
714 end forever.
716 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
718 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
719 it makes less sense with tight coupling for secondary files. But asking
720 the server what the current epoch is might be cheaper on HTTP than on
721 rsync. (Needs to be evaluated)
723 * remove the 0.00 from the verbose overview in the Merged column in the
724 Z row. DONE
726 * write tests that expose the problems of the last few days: cascading
727 client/server roles, tight coupling for secondary RFs, deletes after
728 copies.
730 * Some day we might want to have policy options for the slave:
731 tight/loose/no coupling with upstream for secondary RFs. tight is what
732 we have now. loose would wait until a gap occurs that can be closed.
734 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
736 * revisit all $rfs->[$i+1] places if they now make sense still
738 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
740 * another bug is the fact that the mirror command deletes files before
741 it unhides the index file, thus confusing downstream slaves. We must not
742 delete before unhiding and must delete after unhiding. FIXED.
744 * new complication about the slave that is playing a server role.
745 Currently we mirror from newest to oldest with a hidden temporary file
746 as index. And when one file is finished, we unhide the index file.
747 Imagine the cascading server/slave is dead for a day. It then starts
748 mirroring again with the freshest thing and unhides the freshest index
749 file when it has worked through it. In that moment it exposes a time
750 hole. Because it now works on the second recentfile which is still
751 hidden.
753 We currently do nothing special to converge after such a drop out. At
754 least not intentionally and robustly and thought through.
756 The algorithm we use to seed the next file needs quite a lot of more
757 robustness than it currently has. Something to do with looking at the
758 merged element of the next rf and when it has dropped off, we seed
759 immediately. And if it ramains dropped off, we seed again, of course.
761 Nope, looking from smaller to larger RFS we look at the merged element
762 of this RF and at the minmax/max element of the next RF. If that
763 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
764 seeding it.
766 And we need a public accessor seed and unseed or seeded. But not the mix
767 of public and private stuff that then is used behind the back.
769 And then the secondary* stuff must go.
771 And we must understand what the impact is on the DONE system. Can it go
772 unnoticed that there was a hole? And could the DONE system have decided
773 the hole is covered? This should be testable with three directories where
774 the middle stops working for a while. Done->merge is suspicious, we must
775 stop it from merging non-conflatable neighbors due to broken continuity.
777 FIXED
779 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
781 * Slaven suggests to have the current epoch or the whole current
782 recentfile available from the HTTP server and take it away with
783 keepalive. This direction goes the granularity down to subseconds.
785 We might want to rewrite everything to factor out transport and allow
786 the whole thing to run via HTTP.
788 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
790 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
791 symlink?
793 * smoker on k81 fetching from k75 to verify cascading works. See
794 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
796 * maybe the loop should wait for CHECKSUMS file after every upload. And
797 CPAN.pm needs to deal with timestamps in the future.
799 * do not forget the dirtymark!
801 Text: have a new flag on recentfiles with the meaning: if this
802 changes, you're required to run a full rsync over all the files. The
803 reason why we set it would probably be: some foul happened. we injected
804 files in arbitrary places or didn't inject them although they changed.
805 The content of the flag? Timestamp? The relation between the
806 recentfiles would have to be inheritance from the principal, because any
807 out of band changes would soon later propagate to the next recentfile.
809 By upping the flag often one can easily ruin the slaves.
811 last out of band change? dirtymark?
813 Anyway, this implies that we read a potentially existing recentfile
814 before we write one.
816 And it implies that we have an eventloop that keeps us busy in 2-3
817 cycles, one for current stuff (tight loop) and one for the recentfiles
818 (cascade when principal has changed), one for the old stuff after a
819 dirtymark change.
821 And it implies that the out-of-band change in any of the recentfiles
822 must have a lock on the principal file and there is the place to set the
823 dirtymark.
825 * start a FAQ, especially quick start guide questions. Also to aid those
826 problematic areas where we have no good solution, like the "links"
827 option to rsync.
829 * wish feedback when we are slow.
831 * reduce mccabe
833 * Remove a few DEBUG statements.
835 * The multiple-rrr way of doing things needs a new option to rmirror,
836 like piecemeal or so. Not urgent because after the first pass through,
837 things run smoothely. It's only ugly during the first pass.
839 * I have the suspicion that the code is broken that decides if the
840 neighboring RF needs to be seeded. I fear when too much time has gone
841 between two calls (in our case more than one hour), it would not seed
842 the neighbor. Of course this will never be noticed, so we need a good
843 test for it.
845 * local/localroot confusion: I currently pass both options but one must
846 do.
848 * accounts for early birds on PAUSE rsync daemon.
850 * hardcoded 20 seconds
852 * who mirrors the index? DOING now.
854 * which CPAN mirrors offer rsync?
856 * visit all XXX, visit all _float places
858 * rename the pathdb stuff, it's too confusing. No idea how.
860 * rrr-inotify, backpan, rrr-register
862 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
864 * current bugs: the pathdb seems to get no reset, the seeding of the
865 secondaryttl stuff seems not to have an effect. Have helped myself with
866 a rand(10), need to fix this back. So not checked in. Does the rand
867 thing even help?
869 The rand thing helps. The secondaryttl stuff was in the wrong line,
870 fixed now.
872 The pathdb stuff was because I called either _pathdb or __pathdb on the
873 wrong object. FIXED now.
875 * It's not so beautiful if we never fetch the recentfiles that are not
876 the principal, even if this is correct behaviour. We really do not need
877 them after we have fetched the whole content.
879 OK, we want a switch for that: secondaryttl DONE
881 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
883 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
884 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
885 it lets 40 follow. FIXED
887 * See that the long running process really only updates the principal
888 file unless it has missed a timespan during which something happened. If
889 nothing happened, it must notice even when it misses the timespan. DONE
891 * we must throw away the pathdb when we have reached the end of Z. From
892 that moment we can have a very small pathdb because the only reason for
893 a pathdb is that we know to ignore old records in old files. We won't
894 need this pathdb again before the next full pass over the data is
895 necessary and then we will rebuild it as we go along. DONE
897 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
899 * I think, Done::register_one is doing wrong in that it does not
900 conflate neighboring pieces. The covered() method cannot do this because
901 it has no recent_events array at hand. But register_one has it and could
902 do it and for some reason misses to do it (sometimes).
904 This means that the three tests I just wrote can probably not survive
905 because they test with an already broken Done structure.
907 The art now is to detect how it happens, then to reproduce, then write a
908 test, then fix it.
910 So from the logfile this is what happens: we have a good interval with
911 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
912 top of it. Locally we now mirror F2 and open a new done interval for it.
913 Then we mirror F1 but this time with the timestamp T1b. And when we then
914 try to close the gap, we do not find T1 but instead something older. We
915 should gladly accept this older piece and this would fix this bug.
917 FIXED
919 * bug to fix: when the 1h file changes while rmirror is running, we do
920 correctly sync the new files but never switch to the 6h file but rather
921 stay in a rather quick loop that fetches the 1h file again and again.
923 Is it possible that we initialize a new object? Or does
924 get_remote_recentfile_as_tempfile overwrite something in myself?
926 Want a new option: _runstatusfile => $file which frequently dumps the
927 state of all recentfiles to a file.
929 FIXED
931 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
933 * Todo: now teach update to verify the timestamp is about to write
934 against the previous and use _increase_a_bit if it doesn't comply with
935 strict monotony. DONE
937 * The problem of rounding. So far perl's default precision was
938 sufficient. One day it won't be. FakeFloat has an easy job when it is
939 only reading and other machines have written correctly. But when we want
940 to write a floating point number that is a bit larger than the other
941 one, then we need our own idea of precision.
943 Slaven said: just append a "1". This might be going towards the end of
944 usability too quickly. I'd like something that actually uses the decimal
945 system. Well, appending a 1 also does this but...
947 E.g. we have 1.0. nextup on this architecture is starting with
948 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
949 taken the 1.0000000000000003 and the next user comes and the time tells
950 him 1.0 again. He has to beat my number without stepping over the
951 nextup. This is much less space than I had when I chose 1,2,3.
953 What is also irritating is that nextup is architecture dependent. The
954 128 bit guy must choose very long numbers to fit in between whereas the
955 other one with 16 bit uses larger steps. But then the algorithm is the
956 same for both, so that would be a nice thing.
958 I see two situation where we need this. One is when Time::HiRes returns
959 us a value that is <= the last entry in our recentfile. In this case
960 (let's call it the end-case) we must fill the region between that number
961 and the next higher native floating point number. The other is when we
962 inject an old file into an old recentfile (we would then also set a new
963 dirtymark). We find the integer value already taken and need a slightly
964 different one (let's call it the middle-case). The difference between
965 the two situations is that the next user will want to find something
966 higher than my number in the end-case and something lower than my number
967 in the middle case.
969 So I suggest we give the function both a value and an upper bound and it
970 calculates us a primitive middle. The upper bound in the middle-case is
971 the next integer. The upper bound on the end-case is the nextup floating
972 point number. But the latter poses another problem: if we have occupied
973 the middle m between x and nextup(x), then the nextup(m) will probably
974 not be the same as nextup(x) because some rounding will take place
975 before the nextup is calculated and when the rounding reaches the
976 nextup(x), we will end up at nextup(nextup(x)).
978 So we really need to consider the nextup and the nextdown from there and
979 then the middle and that's the number we may approach asymptotically.
980 Ugly. But DONE.
982 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
984 * consider deprecating the use of RECENT.recent as a symlink. It turns
985 out to need extra hoops with the rsync options and just isn't worth it.
986 Or maybe these extra hoops are needed anyway for the rest of the tree?
987 Nope, can't be the case because not all filesystems support symlinks.
989 But before doing the large step, I'll deprecate the call of
990 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
991 only introduced to resolve RECENT.recent and complicates the routine far
992 beyond what it deserves.
994 DONE. Won't deprecate RECENT.recent, just moved its handling to the
995 supervisor.
997 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
999 * I think it's a bug that the rsync_option links must be set to true in
1000 order to support RECENT.recent and that nobody cares to set it
1001 automatically. Similar for ignore_link_stat_errors.
1003 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1005 * Todo: find all todos together and make a plan what is missing for a
1006 release.
1008 - verifytree or something like that. fsck maybe.
1010 - rersyncrecent, the script itself? What it do?
1012 - a way to only mirror the recentfiles without mirroring the whole
1013 remote system such that people can decide to mirror only partially see
1014 also 2008-08-30. .shadow-xxx directory? this also needed for a
1015 filesystem that is still incomplete and might need the mirrorfiles for
1016 lookup(?)
1018 - long living objects that mirror again and again. Inject something
1019 into ta, see how it goes over to tb.
1021 - how do we continue filling up the DONE system when we use an object
1022 for the second time? "fully covered" and "uptodate" or new terminology.
1024 - overview called on the wrong file should be understandable
1026 - the meta data field that must change when we fake something up so that
1027 the downstream people know they have to re-fetch everything.
1029 - how tolerant are we against missing files upstream? how do we keep
1030 track? there are legitimate cases where we did read upstream index right
1031 before a file got deleted there and then find that file as new and want
1032 it. There are other cases that are not self healing and must be tracked
1033 and bugreported.
1035 - how, exactly, do we have to deal with deletes? With rsync errors?
1037 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
1038 authors) failed: No such file or directory (2)
1040 The file above is a delete in 1h and a new in file 1M and the
1041 delete in the locally running rmirror did not get propagated to the 1M
1042 object. Bug. And the consequence is a standstill.
1044 It seems that a slave that works with a file below the principal needs
1045 to merge things all the way up to get rid of later deletes. Or keep
1046 track of all deletes and skip them later. So we need a trackdeletes.pm
1047 similar to the done.pm?
1049 see also 2008-08-20 about spurious deletes that really have no add
1050 counterpart and yet they are not wrong.
1052 - consider the effect when resyncing the recentfile takes longer than
1053 the time per loop. Then we never rsync any file. We need to diagnose
1054 that and force an increase of that loop time. But when we later are fast
1055 enough again because the net has recovered, then we need to switch back
1056 to original parameters. ERm, no, it's enough to keep syncing at least
1057 one file before refetching an index file.
1059 - remember to verify that no temp files are left lying around and the
1060 signal handler
1062 - status file for not long running jobs that want to track upstream with
1063 a, say, cronjob.
1065 - revisit all XXX _float areas and study Sub::Exporter DONE
1067 - persistent DB even though we just said we do not need it. Just for
1068 extended capabilities and time savings when, for example, upstream
1069 announces a reset and we get new recentfiles and could then limit
1070 ourselves to a subset of files (those that have a changed epoch) in a
1071 first pass and would only then do the loop to verify the rest. Or
1072 something.
1074 * Todo: aggregate files should know their feed and finding the principal
1075 should be done stepwise. (?)
1077 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
1078 left locks around. DONE
1080 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1082 * maybe extend the _overview so that it always says if and where the
1083 last file is in the next file and where the next event in the next rf
1084 would lie. No, don't like this anymore. REJECT
1086 * take the two new redundant tests out again, only the third must
1087 survive. DONE
1089 * Todo: add a sanity check if the merged structure is really pointing to
1090 a different rf and that this different rf is larger. DONE
1092 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1094 * now test, if they are overlapping. And test if there is a file in the
1095 next rf that would fit into this rf's interval.
1097 1h 1222324012.8474 1222322541.7963 0.4086
1098 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
1099 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
1100 1W 1222313218.3626 1221708477.5829 167.9835
1102 I suspect that somebody writes a merged timestamp without having merged
1103 and then somebody else relies on it.
1105 If aggregate is running, the intervals must not be extravagated, if it
1106 is not running, there must not be bounds, the total number of events in
1107 the system must be counted and must be controlled throughout the tests.
1108 That the test required the additional update was probably nonsense,
1109 because aggregate can cut pieces too. FIXED & DONE
1111 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1113 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1114 changed. FIXED
1116 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1118 * Most apparent bug at the moment is that the recentfiles are fetched
1119 too often. Only the principal should be fetched and if it has not
1120 changed, the others should not be refetched. ATM I must admit that I'm
1121 happy that we refetch more often than needed because I can more easily
1122 fix bugs while the thing is running.
1124 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1125 done but the done system does not know about it. The reason for the
1126 failure is not known and we never reach the status uptodate because of
1127 this. We must get over it.
1129 Later it turns out that the origin server had a bug somewhere.
1130 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1131 array of the recentfile one position above. The bug was my own.
1133 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1135 * There is the race condition where the server does a delete and the
1136 slave does not yet know and then tries to download it because he sees
1137 the new. So for this time window we must be more tolerant against
1138 failure. If we cannot download a file, we should just skip it and should
1139 not retry immediately. The whole system should discover the lost thing
1140 later. Keeping track with the DONE system should really be a no brainer.
1142 But there is something more: the whole filesystem is a database and the
1143 recentfiles are one possible representation of it. It's a pretty useful
1144 representation I think that's why I have implemented something around
1145 it. But for strictly local operation it has little value. For local
1146 operation we would much rather have a database. So we would enter every
1147 recentfile reading and every rsync operation and for every file the last
1148 state change and what it leads to. Then we would always ignore older
1149 records without the efforts involved with recentfiles.
1151 The database would have: path,recentepoch,rsyncedon,deletedon
1153 Oh well, not yet clear where this leads to.
1155 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1157 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1158 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1160 * statusfile, maybe only on demand, alone to have a sharp debugging
1161 tool. It is locked and all recentfiles dump themselves into it and we
1162 can build a viewer that lets us know where we stand and what's inside.
1164 * remember: only the principal recentfile needs expiration, all others
1165 shall be expired by principal if it discovers that something has move
1166 upstream.
1168 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1170 * Always check if we stringify to a higher value than in the entry
1171 before. DONE
1173 * And in covered make an additional check if we would be able to see a
1174 numerical difference between the two numbers and if we can't then switch
1175 to a different, more expensive algorithm. Do not want to be caught by
1176 floating surprises. DONE
1178 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1180 * caching has several aspects here: we can cache the interval of the
1181 recentfile which only will change when the mtime of the file changes. We
1182 must re-mirror the recentfile when its ttl has expired. Does have_read
1183 tell you anything? It counts nothing at all. Only the mtime is
1184 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1185 thing the have_mirrored because it is unrelated to the mtime.
1187 * Robustness of floating point calculations! I always thought that the
1188 string calculated by the origin server for the floating representation
1189 of the epoch time is just a string. When we convert it to a number and
1190 later back to a string, the other computer might come to a different
1191 conclusion. This must not happen, we want to preserve it under any
1192 circumstances. I will have to write tests with overlong sequences that
1193 get lost in arithmetic and must see if all still works well. DONE
1195 But one fragile point remains: if one host considers a>b and the other
1196 one considers them == but no eq. To prevent this, we must probably do
1197 some extra homework. DONE
1199 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1201 * the concept of tracking DONE needs an object per recentfile that has
1202 something like these methods:
1204 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1206 covered() register() covered()
1208 The unclear thing is how we translate points in time into intervals. We
1209 could pass a reference to the current recent_events array when running
1210 we_have(xxx) and let the DONE object iterate over it such that it only
1211 has to store a list of intervals that can melt into each other. Ah, even
1212 passing the list together with a list of indexes seems feasiable.
1214 Or maybe ask for the inverted list?
1216 Whenever the complete array is covered by the interval we say we are
1217 fully covered and if the recentfile is not expired, we are uptodate.
1219 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1221 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1223 * need a way to "return" the next entry after the end of a list. When
1224 the caller says "before" or "after" we would like to know if he could
1225 cover that interval/threshold or not because this influences the effect
1226 of a newer timestamp of that recentfile. DONE with $opt{info}.
1228 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1230 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1232 loop implies tixu (?). Nope, something like --statefile decides. Per
1233 default we do ...?
1235 T test, I init, X including removals, U nodirtymark
1237 So we have no concept of dirtymarks, we only trust that since we are
1238 running we have observed everything steadily. But people will not let
1239 this program run forever so we must consider both startup penalty and
1240 book keeping for later runs. We keep this for later. For now we write a
1241 long running mirror that merges several intervals.
1243 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1245 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1246 the intervals!
1248 * rersyncrecent, the script: default to one week. The name of the switch
1249 is --after. Other switches? --loop!
1251 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1253 * need a switch --skip-deletes (?)
1255 * need a switch --enduser that tells us that the whole tempfile
1256 discipline is not needed when there is no downstream user. (?)
1258 Without this switch we cannot have a reasonable recent.pl that just
1259 displays the recent additions. Either we accept to download everything.
1260 Or we download temporary files without the typical rsync protocol
1261 advantages.
1263 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1264 File::Temp::tempdir, this might be a win.
1266 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1268 * apropos missing: we have no push, we never know the downstream
1269 servers. People who know their downstream hosts and want to ascertain
1270 something will want additional methods we have never thought about, like
1271 update or delete a certain file.
1273 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1275 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1276 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1277 difference.
1279 * rsync options: it's a bit of a pain that we usually need several rsync
1280 options, like compress, links, times, checksum and that there is no
1281 reasonable default except the original rsync default. I think wee can
1282 safely assume that the rsync options are shared between all recentfile
1283 instances within one recent tree.
1285 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1287 * deletes: if a delete follows an add quickly enough it may happen that
1288 a downstream mirror did not see the add at all! It seems this needs to
1289 be mentioned somewhere. The point here is that even if the downstream is
1290 never missing the principal timeframe it may encounter a "delete" that
1291 has no complimentary "add" anywhere.
1293 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1295 * I suspect the treat of metadata is incorrect during read or something.
1296 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1297 contained more than 6 hours worth of data. At 06:08 we merged into the
1298 1d file. We need to take snapshots of the 6h file over the course of an
1299 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1300 enough.
1302 Much worse: watching the 1h file: right at the moment (at 06:35) it
1303 covers 1218867584-1219120397 which is 70 hours.
1305 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1306 08:19:44 2008, that is when I checked out last time, so it seems to be
1307 aggregating and never truncating?
1309 No, correct is: it is never truncating; but wrong is: it is aggregating.
1310 It does receive a lot of events from time to time from a larger file.
1311 Somehow a large file gets merged into the small one and because the
1312 "meta/merged" attribute is missing, nobody is paying attention. I
1313 believe that I can fix this by making sure that metadata are honoured
1314 during read. DONE and test adjusted.
1316 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1318 * grand renaming plan
1320 remotebase => remoteroot to fit well with localroot DONE
1321 local_path() => localroot seems to me should already work DONE
1322 recentfile_basename => rfilename no need to stress it has no slash DONE
1324 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1325 kern? It would anyway need a deprecation cycle because it is an
1326 important constructor.
1328 * I like the portability that Data::Serializer brings us but the price
1329 is that some day we might find out that it is slowing us a bit. We'll
1330 see.
1332 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1334 * should we not enter the interval of the principal (or the interval of
1335 the merging file?) in every aggregated/merged file?
1337 * we should aim at a first release and give up on thinking about
1338 sanitizing stuff and zloop. Let's just admit that a full traditional
1339 rsync is the only available sanitizer ATM. Otherwise it's complicated
1340 stuff: sanitizing on the origin server, sanitizing on the slaves,
1341 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1342 get the basics out before this becomes a major cause for mess.
1344 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1346 * On OSes not supporting symlinks we expect that RECENT.recent contains
1347 the contents of the principal recentfile. Actually this is identical on
1348 systems supporting symlinks. Simple, what follows from that is that we
1349 need to keep the serializer in the metadata because we cannot read it
1350 from the filename, doesn't it? Of course not. It's a chicken and egg
1351 problem. This leaves us with the problem to actually parse the
1352 serialized data to find out in which format it is. So who can do the 4
1353 or 5 magics we wanted to support? File::LibMagic?
1355 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1357 * remotebase and recentfile_basename are ugly names. Now that we need a
1358 word for the shortest/principal/driving recentfile too we should do
1359 something about it.
1361 localroot is good. rfile is good. local_path() is bad, local_path($path)
1362 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1363 already deprecated.
1365 Up to now remotebase was the string that described the remote root
1366 directory in rsync notation, like pause.perl.org::authors. And
1367 recentfile_basename was "RECENT-1h.yaml".
1369 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1371 * The test that was added in today's checkin is a good start for a test
1372 of rmirror. We should have more methods in Recent.pm: verify,
1373 addmissingfiles. We should verify the current tree, then rmirror it and
1374 then verifytree the copy. We could then add some arbitrary file and let
1375 it be discovered by addmissingfiles, then rmirror again and then
1376 verifytree the copy again.
1378 Then we could start stealing from csync2 sqlite database [no port to
1379 OSX!] and fill a local DB. And methods to compare the database with the
1380 recentfiles. Our strength is that in principle we could maintain state
1381 with a single float. We have synced up to 1234567890.123456. If the Z
1382 file does not add new files all we have to do is mirror the new ones and
1383 delete the goners.
1385 This makes it clear that we should extend current protocol and declare
1386 that we cheat when we add files too late, just to help the other end
1387 keeping track. Ah yes, that's what was meant when zloop was mentioned
1388 earlier.
1390 Maybe need to revisit File::Mirror to help me with this task.
1392 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1394 * There must be an allow-me-to-truncate flag in every recentfile.
1395 Without it one could construct a sequence of updates winning the locking
1396 battle against the aggregator. Only if an aggregator has managed to
1397 merge data over to the next level, truncating can be allowed. DONE with
1398 accessor merged.
1400 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1402 * We should probably guarantee that no duplicates enter the aggregator
1403 array.
1405 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1407 * To get merge operation faster would need a good benchmark test. What
1408 02 spits out isn't reliable enough and is dominated by many other
1409 things. Between
1411 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1412 Author: Andreas J. Koenig <k@k75.(none)>
1413 Date: Sat Aug 2 07:58:04 2008 +0200
1415 and
1417 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1418 Author: Andreas J. Koenig <k@k75.(none)>
1419 Date: Sat Aug 2 11:40:29 2008 +0200
1421 there was a lot of trying but the effect is hardly measurable with
1422 current tests.
1424 * overhead of connecting seems high. When setting
1425 max_files_per_connection to 1 we see that.
1427 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1429 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1431 25138.2215600014/3600 = 6.98283932222261
1433 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1434 be a bug in the aggregator. FIXED
1436 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1438 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1439 we have it, can we mirror it, mirror it!
1441 I fear this needs a new class which might be called
1442 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1443 and treat them as an entity. I realize that a single recentfile may be
1444 sufficient for certain tasks and that it is handy for the low level
1445 programmer but it is not nice to use. If there is a delete in the 1h
1446 file then the 6h file still contains it. Seekers of the best information
1447 need to combine at least some of the recentfiles most of the time.
1449 There is the place for the Z loop!
1451 But the combination is something to collect in a database, isn't it. Did
1452 csync2 just harrumph?
1454 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1456 * it just occurred to me that hosts in the same mirroring pool could
1457 help out each other even without rewriting the recentfile. Just fetch
1458 the stuff to mirror from several places, bingo. But that's something
1459 that should rather live in a separate package or in rsync directly.
1461 * cronjobs are unsuited because with ntp they would all come at the full
1462 minute and disturb each other. Besides that I'd hate to have a backbone
1463 with more than a few seconds latency.
1465 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1467 * a second rsync server with access control for PAUSE. Port? 873 is the
1468 standard port, let's take 8873.
1470 * if there were a filesystem based on this, it would have a slow access
1471 to inexistent files. It would probably provide wrong readdir (only based
1472 on current content) or also a slow one (based on a recentfile written
1473 after the call). But it would provide fast access to existing files. Or
1474 one would deliberately allow slightly blurred answers based on some
1475 sqlite reflection of the recentfiles.
1477 * todo: write a variant of mirror() that combines two or more
1478 recentfiles and treats them like one
1480 * todo: signal handler to remove the tempfile
1482 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1484 * now that we have the symlink I forgot how it should be used in
1485 practice.
1487 * the z loop: add missing files to Z file. Just append them (instead of
1488 prepending). So one guy prepends something from the Y file from time to
1489 time and another guy appends something rather frequently. Collecting
1490 pond. When Y merges into Z, things get epoch and the collecting pond
1491 gets smaller. What exactly are "missing files"?
1493 take note of current epoch of the alpha file, let's call it the
1494 recent-ts
1496 find all files on disk
1498 remove all files registered in the recentworld up to recent-ts
1500 remove all files that have been deleted after recent-ts according to
1501 recentworld
1503 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1505 * rersyncrecent might be a cronjob with a (locked) state file which
1506 contains things like after and maybe last z sync or such?
1508 rrr-mirror might be an alternative name but how would we justify the
1509 three Rs when there is no Re-Rsync-Recent?
1511 With the --loop parameter it is an endless loop, without it is no loop.
1512 At least this is simple.
1514 * todo: new accssor z-interval specifies how often the Z file is updated
1515 against the filesystem. We probably want no epoch stamp on these
1516 entries. And we want to be able to filter the entries (e.g. no
1517 by-modules and by-category tree)
1519 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1521 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1522 into Z. We do not need the other files filled up with filesystem
1523 contents though.
1525 * need interface to query for a file in order to NOT call update on
1526 PAUSE a second time within a short time.
1528 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1530 * recommended update interval? Makes no sense, is different for
1531 different users.
1533 * Moosify
1535 Local Variables:
1536 mode: change-log
1537 change-log-default-name: "Todo"
1538 tab-width: 2
1539 left-margin: 2
1540 End: