| blob | 184c44e7b524cc7a7558fa37bda09111c326d524 |
1 2011-02-17 Andreas J. Koenig <andk@cpan.org>
3 * ALERT rrr: we have a piece of code somewhere in the mirror()
4 subroutine that reads the raw YAML file with normal open() and cuts it
5 off for efficiency, then feeds it to yaml::load in order to get some
6 metadata. Reading the thing with the whole tail of the recent array
7 kills the whole idea. The whole thing will not work with JSON.
9 So we must invent a protocol 3 that works with two index files.
11 2011-02-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
13 * I think interval guaranteed with the $last_aggregate_call
14 variable (both in fsck and server) need to be fixed with a $recc
15 variable to something like 60 seconds or even less.
17 * minor bug: fsck just added the lockfile to the index which should be
18 considered bookkeeping.
20 * speed: a large directory remove that is done by the kernel in 1
21 seconds triggers minutes of bookkeeping (07:56:42 - 08:01:10 for ~3000
22 files, all due to Schlemiehl again. Want to do some collecting of
23 pending ops before actually writing to the RECENT files. Lock!
25 Repeating the timing with 2600 files and it took 21:28:08 - 21:35:27.
27 Now having rewritten the loop to use batch_update(): 4383 file removed
28 21:43:37 - 21:44:12. From 7:30 to 0:35 while doing 60% more work that's
29 a joy.
31 * lockdirectory expiration? server died and blocked fsck for so long.
33 * At the moment rrr-init is not needed anymore, rrr-server can be
34 started and followed by rrr-fsck. Still stupid but not the main
35 showstopper.
37 2011-02-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
39 * IN_Q_OVERFLOW needs to trigger an fsck, likewise the entry into the
40 server.
42 2011-01-30 Andreas J. Koenig <andk@cpan.org>
44 * rrr-init and rrr-server need to be used such:
46 init
47 server
48 init again
50 Stupid but no showstopper ATM.
52 2010-10-26 Andreas J. Koenig <andk@cpan.org>
54 * forking slave is done, needs more testing.
56 2010-07-06 Andreas J. Koenig <andk@cpan.org>
58 * The two next most important goals are to get a useable rrr-server and
59 a forking slave implementation that uses virtually no memory after the
60 first pass through. Does it need the status file?
62 * rrr-init now being fast we can now remove the alpha alerts. Done.
64 2010-06-16 Andreas J. Koenig <andk@cpan.org>
66 * rrr-init seems to be sssslllloooowwww. At 3:49 I started it for the
67 video tree. After twelve minutes the file has reached a size of 500k. It
68 is locking and rewriting the YAML after every file. Wasn't there an
69 unlocked variant? For "init" one would not expect locked correctness and
70 a rewrite after every item. Given that this tends to become slower with
71 every file it is really unacceptable.
73 Reminds me of rrr-dirtyupdate (the former name of rrr-update). It seems
74 the code was not dirty enough and that was the reason why it was so
75 slow. And rrr-update does not have a --dirty switch, so it looks like we
76 have no way to force dirty and fast operation yet but we need it for
77 init.
79 Irritating the bootstrapping race: the order {(1) fill initial
80 array, (2) start server} leaves the need for an fsck to cover the time
81 between 1 and 2. Given that regular fsck is needed in any case the price
82 is not high but still doesn't give a warm feeling. Maybe starting the
83 server will by itself do an fsck?
85 2010-06-15 Andreas J. Koenig <andk@cpan.org>
87 * remove the wild alpha alerts and say, it is alpha but parts of it are
88 already working very well or so.
90 2009-12-10 Andreas J. Koenig <andk@cpan.org>
92 * Prominent request for statusfile to safe memory consumption. I'd like
93 to work backwards: get a verification program that verifies how well in
94 sync two nodes are without using frmr itself. Pretend that the forking
95 agent already works and can be interrupted with ^C. Then let it die at
96 random and restart at random and see the memory consumption and the
97 failure rate over time. Compare it with the curve we have now.
99 Of course we also invented the status file as a sharp debugging tool.
100 I'd like to see it in action. See also: F:R:M:Recent::thaw().
102 2009-06-22 Andreas J. Koenig <andk@cpan.org>
104 * Todo: rrr-init to generate some index files so that rrr-server can be
105 started.
107 2009-05-02 Andreas J. Koenig <andk@cpan.org>
109 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
110 so we can use it for normal update with a single file. And add a cronjob
111 that runs rrr-fsck so we get a reminder when something is not working
112 as it should.
114 2009-04-27 Andreas J. Koenig <andk@cpan.org>
116 * Todo: I still see more link_stat errors than expected. I come back to
117 the former suspicion that we do not do necessary bookkeeping on deleted
118 files so that when we reach the "new" event on a meanwhile deleted file
119 we try to mirror it although we could know better. It's just noise, no
120 harm but noise is irritating.
122 * Todo: more experimenting with runstatusfile to get it official
124 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
125 parse the error.
127 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
129 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
131 And that ignore_link_stat_error is propagated in sparse_clone.
133 DONE
135 For the record: the bug was that ignore_link_stat_error was lost during
136 sparse_clone. And then the second bug was that it did not default to
137 true. Both are now fixed.
139 2009-04-25 Andreas J. Koenig <andk@cpan.org>
141 * Todo: fill rrr program with life
143 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
145 * Minor bug: I see complaints about files not existing like
146 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
147 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
149 In the RECENT files these are delete events but of course also new
150 events in older recent files.
152 I think we do not properly remember deletes when we skip-deletes and so
153 do not filter them out when we later see the "new" event.
155 Yes, but that what ignore_link_stat_error is for, we know about this
156 race condition. We must make ignore_link_stat_error default to true.
158 DONE
160 * bug in Done.pm failing to merge two fields into one when a third is
161 present. FIXED and accompanied with new test.
163 2009-04-24 Andreas J. Koenig <andk@cpan.org>
165 * Todo: keep the index clean and run some kind of nondestructive fsck 4
166 times a day.
168 * who is our backbone?
170 http://cpan.cpantesters.org/authors/02STAMP # barbie
171 http://cpan.solfo.com/authors/02STAMP # abh
172 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
174 2009-04-17 Andreas J. Koenig <andk@cpan.org>
176 * the changelog helper I included in the Makefile release memo does not
177 show tags and is much less useful than I thought. gitk is probably much
178 more convenient.
180 2009-04-16 Andreas J. Koenig <andk@cpan.org>
182 * ABH writes:
184 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
185 #### Ydr_.yaml ... DONE
186 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
187 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
188 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
189 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
190 #### gone already at cpan-pause.pl line 0
192 Running without skip-deletes now but cannot reproduce.
194 2009-04-15 Andreas J. Koenig <andk@cpan.org>
196 * consider state directory such we can restart after a ^C or come again
197 from a cronjob.
199 Reuse _runstatusfile: write atomically, see how we can use the dump to
200 start running again. Do not leak and remove globs before writing. Do not
201 only dump one rf, dump the whole complex.
203 * consider the effect of ^C. Is it safe?
205 * want to use Perl::Version to up the versions on every release. Only if
206 things have changed, of course. Does this actually work?
208 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
209 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
211 Must do the bookkeeping myself when this should happen, maybe based on git?
213 I think I prefer the same version number for all pm files. But one also
214 needs a routine for setting them. Ahh:
216 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
218 NEEDSMOREWORK
220 * why indexer wrong?
222 cpan[2]> m /Mirror::Recent/
223 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
224 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
225 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
226 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
227 4 items found
229 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
231 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
233 I just tried MM 6.5102 and it again produces only a provides for Recent,
234 not for the other modules.
236 My own fault. FIXED in the Makefile.PL.
238 2009-04-13 Andreas J. Koenig <andk@cpan.org>
240 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
242 * is it true (as stated at
243 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
244 that rsync can lead to errors when upstream changes before client sync
245 has completed? Can we see the error when we run this on upstream:
247 % perl -e '
248 use Time::HiRes qw(time);
249 while (){
250 open my $fh, ">", "changingfile.txt.new" or die;
251 print $fh (time."\n") x 1000000;
252 close $fh;
253 rename "changingfile.txt.new", "changingfile.txt" or die;
254 }
255 '
257 And on the receiving end:
259 % while true; do
260 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
261 sleep 1;
262 done
264 I cannot get it to fail. Apparently it is sufficient when upstream
265 always writes atomically (which of course is mandatory)?
267 SENT to mailing list Apr 28.
269 * from ABH:
271 > https://fedorahosted.org/InstantMirror/
272 > https://www.redhat.com/mailman/listinfo/instantmirror-list
273 > irc.freenode.net channel #instantmirror
275 A cool name for a project. Inspires me to write a few sentences about
276 bittorrent's role in the grand picture.
278 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
279 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
281 2009-04-12 Andreas J. Koenig <andk@cpan.org>
283 * Interesting last minute bug during real download testing: the output
284 isn't as pretty anymore, it seems that more work is being done than
285 needed.
287 Introducing a highlevel _alluptodate method helped a lot in debugging
288 this.
290 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
291 different on 1h file than on the other files. This lead to frequent
292 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
293 issue.
295 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
297 * Todo: try if we can be faster with a native float package. I'm really
298 glad to have a machine exposing perl floating point bugs, it forced me
299 to retract this item from the changes file: "several speedups in the
300 fakebigfloat code" before I get cpan testers fails.
302 2009-04-11 Andreas J. Koenig <andk@cpan.org>
304 * Bug: when the dirtymark changes upstream, then the downstream server
305 notices it immediately and mirrors the "1h" file but then leaves the
306 rmirror loop and continues with "6h" on the next round through. This
307 behaviour goes on until it reaches the Z file. Because we eat the Z file
308 piecemeal we leave the loop after a while but after that everybody seems
309 to have forgotten that there is some work left to be done. Bad, bad,
310 bad.
312 ETOOCONFUSING better name?
314 Recentfile::get_remote_recentfile_as_tempfile OK
315 Recentfile::resolve_recentfilename split_rfilename DONE
317 Recent::_principal_recentfile_object _principal_recentfile_fromremote
318 Recent::principal_recentfile OK
319 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
321 some void, some not, some not void but called in void context.
323 2009-04-10 Andreas J. Koenig <andk@cpan.org>
325 * Bug? When the dirtymark changes then the second tier hosts quickly
326 reset their DONE state and restart mirroring. But then they mirror
327 potentially outdated index files with inconsistent dirtymark because the
328 first tier box may be a bit behind on the large recentfiles.
330 This means we need a brake when iterating over the recentfiles that
331 refuses to use a recentfile with the wrong dirtymark.
333 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
335 * install the bin/* files and talk about rrr-overview somewhere.
337 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
338 created outside the target tree.
340 * Barbie asks for better logging capabilities
342 2009-03-29 Andreas J. Koenig <andk@cpan.org>
344 * known bug: cannot configure the Recentfile objects to keep delete
345 events. How to call the accessor for that? keep_delete_objects_forever?
347 Sounds acceptable. FIXED (but untested).
349 * known bug: keeps temporary index files lying around.
351 2009-03-24 Andreas J. Koenig <andk@cpan.org>
353 * Some equivalent for
355 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
357 ? DONE
359 2009-03-22 Andreas J. Koenig <andk@cpan.org>
361 * k81 is again client of k75 with different parameters.
363 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
364 about it all the time and nobody reacts accordingly; (2) lots of temp
365 files get created and not removed; culprit the new call to
366 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
367 caller has to remove the tempfile after use.
369 Need the drawing board.
371 PARTLY FIXED: retracting the idea to call
372 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
373 and unseed such that uptodate now means "we have mirrored this
374 recentfile's bunch of files" and seeded means "somebody believes the
375 index file for this rf needs be refreshed". This seems to work now.
377 Still collecting temp files that nobody cares to depollute.
379 2009-03-21 Andreas J. Koenig <andk@cpan.org>
381 * Bug?: should it be harder than it is atm to set the timestamp to the
382 future? YES, FIXED
384 * bug with native integers:
386 -
387 epoch: 997872011
388 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
389 type: new
390 -
391 epoch: 1195248431
392 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
393 type: new
395 Native integer broke when native math was turned off. FIXED
397 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
398 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
399 in this area.
401 Yes, records out of order:
403 447003 -
404 447004 epoch: 995885533
405 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
406 447006 type: new
407 447007 -
408 447008 epoch: 995890358
409 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
410 447010 type: new
411 447011 -
412 447012 epoch: 995892221
413 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
414 447014 type: new
416 FIXED with sanity check and later with the integer fix.
418 * Bug: want the index files in a .recent directory
420 * Bug: lots of dot files are not deleted in time
422 * possible test case: can a delete change the timestamp? This would
423 probably break the order of events.
425 2009-03-20 Andreas J. Koenig <andk@cpan.org>
427 * 1233701831.34486 what's so special about this number/string? It
428 belongs to
429 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
430 atm lives in Y,Q, and Z.
432 It is the first entry after id/--skip-locking which has timestamp
433 1234164228.11325 which represents a file that doesn't exist anymore.
435 * another bug:
437 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
438 _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
439 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
440 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
441 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
442 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
443 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
444 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
445 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
446 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
447 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
448 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
449 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
451 FIXED, it was the "--skip-locking" file where manual intervention was participating
453 * bug on the mirroring slave: when the dirtymark gets increased we
454 probably do not reset the done intervals. The mirrorer stays within
455 tight bounds where it tries to sync with upstream and never seems to
456 finish. In the debugging state file I see lots of identical intervals
457 that do not get collapsed. When I restart the mirrorer it dies with:
459 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
460 _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
461 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
462 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
463 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
464 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
465 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
466 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
467 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
468 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
470 and debugging stands at
472 At: before
473 Brfinterval: Z
474 Ci: 227
475 Dre+1:
476 epoch: 1237400802.5789
477 path: id/J/JO/JOHND/CHECKSUMS
478 type: new
479 Dre-0:
480 epoch: 1237400807.97514
481 path: id/M/MI/MIYAGAWA/CHECKSUMS
482 type: new
483 Dre-1:
484 epoch: 1237400817.94363
485 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
486 type: new
487 Eintervals:
488 -
489 - 900644040
490 - 900644040
491 - []
493 and it is reproducable.
495 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
496 pause has a fresh one.
498 FIXED, it was the third anded term in each of the ifs in the IV block in
499 _register_one: with that we make sure that we do not stamp on valuable
500 interval data.
502 2009-03-17 Andreas J. Koenig <andk@cpan.org>
504 * done: verified the existence of the floating point bug in bleadperl
505 and verified that switching from YAML::Syck to YAML::XS does not resolve
506 it.
508 BTW, the switch was doable with
510 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
512 and should be considered as a separate TODO
514 * todo: integrate a dirty update with two aggregate calls before
515 unlocking for frictionless dirtying DONE
517 * todo: start the second rsync daemon on pause DONE
519 * todo: move index files to .recent: this cannot simply be done by
520 setting filenameroot to .recent/RECENT. Other parts of the modules rely
521 on the fact that dirname(recentfile) is the root of the mirrored tree.
523 2009-03-16 Andreas J. Koenig <andk@cpan.org>
525 * What was the resolution of the mirror.pl delete hook bug? Do we call
526 the delete hook when pause removes a file from MUIR?
528 * Today on pause: Updating 2a13fba..29f284d and installing it for
529 /usr/local/perl-5.10.0{,-RC2}
531 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
532 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
534 2009-03-15 Andreas J. Koenig <andk@cpan.org>
536 * currently recent_events has the side effect of setting dirtymark
537 because it forces a read on the file. That should be transparent, so
538 that the dirtymark call always forces a cache-able(?) read.
540 * The bug below is -- after a lot of trying -- not reproducible on a
541 small script, only in the large test script. The closest to the output
542 below was:
544 #!perl
545 use strict;
546 use Devel::Peek;
547 my $x = "01237123229.8814";
548 my($l,$r);
549 for ($l,$r) {
550 $_ = "x"x34;
551 }
552 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
553 $r = int $r;
554 $l = "1237123231.22458";
555 $r = "1237123231.22458";
556 1 if $l/1.1;
557 Devel::Peek::Dump $l;
558 Devel::Peek::Dump $r;
559 Devel::Peek::Dump $x = $l <=> $r;
560 die "BROKE" if $x;
561 __END__
563 The checked in state at c404a85 fails the test with my
564 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
565 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
566 in the test script always consistent.
568 * Todo: write a test that inserts a second dirty file with an already
569 existing timestamp. DONE
571 * Bug in perl 5.10 on my 64bit box:
573 DB<98> Devel::Peek::Dump $l
574 SV = PVMG(0x19e0450) at 0x142a550
575 REFCNT = 2
576 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
577 IV = 0
578 NV = 1237123231.22458
579 PV = 0x194ce70 "1237123231.22458"\0
580 CUR = 16
581 LEN = 40
583 DB<99> Devel::Peek::Dump $r
584 SV = PVMG(0x19e0240) at 0x142a3e8
585 REFCNT = 2
586 FLAGS = (PADMY,POK,pPOK)
587 IV = 1237123229
588 NV = 1237123229.8814
589 PV = 0x19ff900 "1237123231.22458"\0
590 CUR = 16
591 LEN = 40
593 DB<100> Devel::Peek::Dump $l <=> $r
594 SV = IV(0x19ea6e8) at 0x19ea6f0
595 REFCNT = 1
596 FLAGS = (PADTMP,IOK,pIOK)
597 IV = -1
599 DB<101> Devel::Peek::Dump $l
600 SV = PVMG(0x19e0450) at 0x142a550
601 REFCNT = 2
602 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
603 IV = 0
604 NV = 1237123231.22458
605 PV = 0x194ce70 "1237123231.22458"\0
606 CUR = 16
607 LEN = 40
609 DB<102> Devel::Peek::Dump $r
610 SV = PVMG(0x19e0240) at 0x142a3e8
611 REFCNT = 2
612 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
613 IV = 1237123231
614 NV = 1237123231.22458
615 PV = 0x19ff900 "1237123231.22458"\0
616 CUR = 16
617 LEN = 40
619 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
621 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
623 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
625 * $obj->merge ($other) needs to learn about equal epoch which may happen
626 since dirty_epoch intruded.
628 * Wontfix anytime soon: I think we currently do not support mkdir. Only
629 files!
631 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
633 * Todo: continue working on update(...,$dirty_epoch). It must be
634 followed by a fast_aggregate! DONE
636 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
638 * maybe we need a closest_entry or fitting_interval or something like
639 that. We want to merge an event into the middle of some recentfile.
640 First we do not know which file, then we do not know where to lock,
641 where to enter the new item, when and where to correct the dirtymark.
643 So my thought is we should first find which file.
645 Another part of my brain answers: what would happen if we would enter
646 the new file into the smallest file just like an ordinary new event,
647 just as an old event?
649 (1) we would write a duplicate timestamp? No, this would be easy to
650 avoid
652 (2) we would make the file large quickly? Yes, but so what? We are
653 changing the dirtymark, so are willing to disturb the downstream hosts.
655 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
657 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
659 leak!
661 https://rt.cpan.org/Ticket/Display.html?id=41199
663 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
665 Strangest output being HAKANARDO who managed to upload
667 Here is a better oneliner that includes also the first line of each
668 finding:
670 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
672 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
674 * memory leak in the syncher? It currently weighs 100M.
676 Update 2008-11-02:
678 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
681 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
683 * lookup by epoch and by path and use this ability on the pause to never
684 again register a file twice that doesn't need it. Let's call it
685 contains().
687 * after the dirtymark is done: fill up recentfiles with fake (historic)
688 entries; fill up with individual corrections; algorithm maybe to be done
689 with bigfloat so that we can always place something in the middle
690 between two entries. Before we must switch to bigfloat we could try to
691 use Data::Float::nextup to get the.
693 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
694 directly.
696 * dirtymark now lives in Recentfile, needs to be used in rmirror.
698 * find out why the downloader died after a couple of hours without a net
699 connection. Write a test that survives the not-existence of the other
700 end forever.
702 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
704 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
705 it makes less sense with tight coupling for secondary files. But asking
706 the server what the current epoch is might be cheaper on HTTP than on
707 rsync. (Needs to be evaluated)
709 * remove the 0.00 from the verbose overview in the Merged column in the
710 Z row. DONE
712 * write tests that expose the problems of the last few days: cascading
713 client/server roles, tight coupling for secondary RFs, deletes after
714 copies.
716 * Some day we might want to have policy options for the slave:
717 tight/loose/no coupling with upstream for secondary RFs. tight is what
718 we have now. loose would wait until a gap occurs that can be closed.
720 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
722 * revisit all $rfs->[$i+1] places if they now make sense still
724 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
726 * another bug is the fact that the mirror command deletes files before
727 it unhides the index file, thus confusing downstream slaves. We must not
728 delete before unhiding and must delete after unhiding. FIXED.
730 * new complication about the slave that is playing a server role.
731 Currently we mirror from newest to oldest with a hidden temporary file
732 as index. And when one file is finished, we unhide the index file.
733 Imagine the cascading server/slave is dead for a day. It then starts
734 mirroring again with the freshest thing and unhides the freshest index
735 file when it has worked through it. In that moment it exposes a time
736 hole. Because it now works on the second recentfile which is still
737 hidden.
739 We currently do nothing special to converge after such a drop out. At
740 least not intentionally and robustly and thought through.
742 The algorithm we use to seed the next file needs quite a lot of more
743 robustness than it currently has. Something to do with looking at the
744 merged element of the next rf and when it has dropped off, we seed
745 immediately. And if it ramains dropped off, we seed again, of course.
747 Nope, looking from smaller to larger RFS we look at the merged element
748 of this RF and at the minmax/max element of the next RF. If that
749 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
750 seeding it.
752 And we need a public accessor seed and unseed or seeded. But not the mix
753 of public and private stuff that then is used behind the back.
755 And then the secondary* stuff must go.
757 And we must understand what the impact is on the DONE system. Can it go
758 unnoticed that there was a hole? And could the DONE system have decided
759 the hole is covered? This should be testable with three directories where
760 the middle stops working for a while. Done->merge is suspicious, we must
761 stop it from merging non-conflatable neighbors due to broken continuity.
763 FIXED
765 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
767 * Slaven suggests to have the current epoch or the whole current
768 recentfile available from the HTTP server and take it away with
769 keepalive. This direction goes the granularity down to subseconds.
771 We might want to rewrite everything to factor out transport and allow
772 the whole thing to run via HTTP.
774 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
776 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
777 symlink?
779 * smoker on k81 fetching from k75 to verify cascading works. See
780 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
782 * maybe the loop should wait for CHECKSUMS file after every upload. And
783 CPAN.pm needs to deal with timestamps in the future.
785 * do not forget the dirtymark!
787 Text: have a new flag on recentfiles with the meaning: if this
788 changes, you're required to run a full rsync over all the files. The
789 reason why we set it would probably be: some foul happened. we injected
790 files in arbitrary places or didn't inject them although they changed.
791 The content of the flag? Timestamp? The relation between the
792 recentfiles would have to be inheritance from the principal, because any
793 out of band changes would soon later propagate to the next recentfile.
795 By upping the flag often one can easily ruin the slaves.
797 last out of band change? dirtymark?
799 Anyway, this implies that we read a potentially existing recentfile
800 before we write one.
802 And it implies that we have an eventloop that keeps us busy in 2-3
803 cycles, one for current stuff (tight loop) and one for the recentfiles
804 (cascade when principal has changed), one for the old stuff after a
805 dirtymark change.
807 And it implies that the out-of-band change in any of the recentfiles
808 must have a lock on the principal file and there is the place to set the
809 dirtymark.
811 * start a FAQ, especially quick start guide questions. Also to aid those
812 problematic areas where we have no good solution, like the "links"
813 option to rsync.
815 * wish feedback when we are slow.
817 * reduce mccabe
819 * Remove a few DEBUG statements.
821 * The multiple-rrr way of doing things needs a new option to rmirror,
822 like piecemeal or so. Not urgent because after the first pass through,
823 things run smoothely. It's only ugly during the first pass.
825 * I have the suspicion that the code is broken that decides if the
826 neighboring RF needs to be seeded. I fear when too much time has gone
827 between two calls (in our case more than one hour), it would not seed
828 the neighbor. Of course this will never be noticed, so we need a good
829 test for it.
831 * local/localroot confusion: I currently pass both options but one must
832 do.
834 * accounts for early birds on PAUSE rsync daemon.
836 * hardcoded 20 seconds
838 * who mirrors the index? DOING now.
840 * which CPAN mirrors offer rsync?
842 * visit all XXX, visit all _float places
844 * rename the pathdb stuff, it's too confusing. No idea how.
846 * rrr-inotify, backpan, rrr-register
848 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
850 * current bugs: the pathdb seems to get no reset, the seeding of the
851 secondaryttl stuff seems not to have an effect. Have helped myself with
852 a rand(10), need to fix this back. So not checked in. Does the rand
853 thing even help?
855 The rand thing helps. The secondaryttl stuff was in the wrong line,
856 fixed now.
858 The pathdb stuff was because I called either _pathdb or __pathdb on the
859 wrong object. FIXED now.
861 * It's not so beautiful if we never fetch the recentfiles that are not
862 the principal, even if this is correct behaviour. We really do not need
863 them after we have fetched the whole content.
865 OK, we want a switch for that: secondaryttl DONE
867 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
869 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
870 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
871 it lets 40 follow. FIXED
873 * See that the long running process really only updates the principal
874 file unless it has missed a timespan during which something happened. If
875 nothing happened, it must notice even when it misses the timespan. DONE
877 * we must throw away the pathdb when we have reached the end of Z. From
878 that moment we can have a very small pathdb because the only reason for
879 a pathdb is that we know to ignore old records in old files. We won't
880 need this pathdb again before the next full pass over the data is
881 necessary and then we will rebuild it as we go along. DONE
883 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
885 * I think, Done::register_one is doing wrong in that it does not
886 conflate neighboring pieces. The covered() method cannot do this because
887 it has no recent_events array at hand. But register_one has it and could
888 do it and for some reason misses to do it (sometimes).
890 This means that the three tests I just wrote can probably not survive
891 because they test with an already broken Done structure.
893 The art now is to detect how it happens, then to reproduce, then write a
894 test, then fix it.
896 So from the logfile this is what happens: we have a good interval with
897 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
898 top of it. Locally we now mirror F2 and open a new done interval for it.
899 Then we mirror F1 but this time with the timestamp T1b. And when we then
900 try to close the gap, we do not find T1 but instead something older. We
901 should gladly accept this older piece and this would fix this bug.
903 FIXED
905 * bug to fix: when the 1h file changes while rmirror is running, we do
906 correctly sync the new files but never switch to the 6h file but rather
907 stay in a rather quick loop that fetches the 1h file again and again.
909 Is it possible that we initialize a new object? Or does
910 get_remote_recentfile_as_tempfile overwrite something in myself?
912 Want a new option: _runstatusfile => $file which frequently dumps the
913 state of all recentfiles to a file.
915 FIXED
917 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
919 * Todo: now teach update to verify the timestamp is about to write
920 against the previous and use _increase_a_bit if it doesn't comply with
921 strict monotony. DONE
923 * The problem of rounding. So far perl's default precision was
924 sufficient. One day it won't be. FakeFloat has an easy job when it is
925 only reading and other machines have written correctly. But when we want
926 to write a floating point number that is a bit larger than the other
927 one, then we need our own idea of precision.
929 Slaven said: just append a "1". This might be going towards the end of
930 usability too quickly. I'd like something that actually uses the decimal
931 system. Well, appending a 1 also does this but...
933 E.g. we have 1.0. nextup on this architecture is starting with
934 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
935 taken the 1.0000000000000003 and the next user comes and the time tells
936 him 1.0 again. He has to beat my number without stepping over the
937 nextup. This is much less space than I had when I chose 1,2,3.
939 What is also irritating is that nextup is architecture dependent. The
940 128 bit guy must choose very long numbers to fit in between whereas the
941 other one with 16 bit uses larger steps. But then the algorithm is the
942 same for both, so that would be a nice thing.
944 I see two situation where we need this. One is when Time::HiRes returns
945 us a value that is <= the last entry in our recentfile. In this case
946 (let's call it the end-case) we must fill the region between that number
947 and the next higher native floating point number. The other is when we
948 inject an old file into an old recentfile (we would then also set a new
949 dirtymark). We find the integer value already taken and need a slightly
950 different one (let's call it the middle-case). The difference between
951 the two situations is that the next user will want to find something
952 higher than my number in the end-case and something lower than my number
953 in the middle case.
955 So I suggest we give the function both a value and an upper bound and it
956 calculates us a primitive middle. The upper bound in the middle-case is
957 the next integer. The upper bound on the end-case is the nextup floating
958 point number. But the latter poses another problem: if we have occupied
959 the middle m between x and nextup(x), then the nextup(m) will probably
960 not be the same as nextup(x) because some rounding will take place
961 before the nextup is calculated and when the rounding reaches the
962 nextup(x), we will end up at nextup(nextup(x)).
964 So we really need to consider the nextup and the nextdown from there and
965 then the middle and that's the number we may approach asymptotically.
966 Ugly. But DONE.
968 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
970 * consider deprecating the use of RECENT.recent as a symlink. It turns
971 out to need extra hoops with the rsync options and just isn't worth it.
972 Or maybe these extra hoops are needed anyway for the rest of the tree?
973 Nope, can't be the case because not all filesystems support symlinks.
975 But before doing the large step, I'll deprecate the call of
976 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
977 only introduced to resolve RECENT.recent and complicates the routine far
978 beyond what it deserves.
980 DONE. Won't deprecate RECENT.recent, just moved its handling to the
981 supervisor.
983 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
985 * I think it's a bug that the rsync_option links must be set to true in
986 order to support RECENT.recent and that nobody cares to set it
987 automatically. Similar for ignore_link_stat_errors.
989 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
991 * Todo: find all todos together and make a plan what is missing for a
992 release.
994 - verifytree or something like that. fsck maybe.
996 - rersyncrecent, the script itself? What it do?
998 - a way to only mirror the recentfiles without mirroring the whole
999 remote system such that people can decide to mirror only partially see
1000 also 2008-08-30. .shadow-xxx directory? this also needed for a
1001 filesystem that is still incomplete and might need the mirrorfiles for
1002 lookup(?)
1004 - long living objects that mirror again and again. Inject something
1005 into ta, see how it goes over to tb.
1007 - how do we continue filling up the DONE system when we use an object
1008 for the second time? "fully covered" and "uptodate" or new terminology.
1010 - overview called on the wrong file should be understandable
1012 - the meta data field that must change when we fake something up so that
1013 the downstream people know they have to re-fetch everything.
1015 - how tolerant are we against missing files upstream? how do we keep
1016 track? there are legitimate cases where we did read upstream index right
1017 before a file got deleted there and then find that file as new and want
1018 it. There are other cases that are not self healing and must be tracked
1019 and bugreported.
1021 - how, exactly, do we have to deal with deletes? With rsync errors?
1023 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
1024 authors) failed: No such file or directory (2)
1026 The file above is a delete in 1h and a new in file 1M and the
1027 delete in the locally running rmirror did not get propagated to the 1M
1028 object. Bug. And the consequence is a standstill.
1030 It seems that a slave that works with a file below the principal needs
1031 to merge things all the way up to get rid of later deletes. Or keep
1032 track of all deletes and skip them later. So we need a trackdeletes.pm
1033 similar to the done.pm?
1035 see also 2008-08-20 about spurious deletes that really have no add
1036 counterpart and yet they are not wrong.
1038 - consider the effect when resyncing the recentfile takes longer than
1039 the time per loop. Then we never rsync any file. We need to diagnose
1040 that and force an increase of that loop time. But when we later are fast
1041 enough again because the net has recovered, then we need to switch back
1042 to original parameters. ERm, no, it's enough to keep syncing at least
1043 one file before refetching an index file.
1045 - remember to verify that no temp files are left lying around and the
1046 signal handler
1048 - status file for not long running jobs that want to track upstream with
1049 a, say, cronjob.
1051 - revisit all XXX _float areas and study Sub::Exporter DONE
1053 - persistent DB even though we just said we do not need it. Just for
1054 extended capabilities and time savings when, for example, upstream
1055 announces a reset and we get new recentfiles and could then limit
1056 ourselves to a subset of files (those that have a changed epoch) in a
1057 first pass and would only then do the loop to verify the rest. Or
1058 something.
1060 * Todo: aggregate files should know their feed and finding the principal
1061 should be done stepwise. (?)
1063 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
1064 left locks around. DONE
1066 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1068 * maybe extend the _overview so that it always says if and where the
1069 last file is in the next file and where the next event in the next rf
1070 would lie. No, don't like this anymore. REJECT
1072 * take the two new redundant tests out again, only the third must
1073 survive. DONE
1075 * Todo: add a sanity check if the merged structure is really pointing to
1076 a different rf and that this different rf is larger. DONE
1078 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1080 * now test, if they are overlapping. And test if there is a file in the
1081 next rf that would fit into this rf's interval.
1083 1h 1222324012.8474 1222322541.7963 0.4086
1084 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
1085 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
1086 1W 1222313218.3626 1221708477.5829 167.9835
1088 I suspect that somebody writes a merged timestamp without having merged
1089 and then somebody else relies on it.
1091 If aggregate is running, the intervals must not be extravagated, if it
1092 is not running, there must not be bounds, the total number of events in
1093 the system must be counted and must be controlled throughout the tests.
1094 That the test required the additional update was probably nonsense,
1095 because aggregate can cut pieces too. FIXED & DONE
1097 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1099 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1100 changed. FIXED
1102 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1104 * Most apparent bug at the moment is that the recentfiles are fetched
1105 too often. Only the principal should be fetched and if it has not
1106 changed, the others should not be refetched. ATM I must admit that I'm
1107 happy that we refetch more often than needed because I can more easily
1108 fix bugs while the thing is running.
1110 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1111 done but the done system does not know about it. The reason for the
1112 failure is not known and we never reach the status uptodate because of
1113 this. We must get over it.
1115 Later it turns out that the origin server had a bug somewhere.
1116 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1117 array of the recentfile one position above. The bug was my own.
1119 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1121 * There is the race condition where the server does a delete and the
1122 slave does not yet know and then tries to download it because he sees
1123 the new. So for this time window we must be more tolerant against
1124 failure. If we cannot download a file, we should just skip it and should
1125 not retry immediately. The whole system should discover the lost thing
1126 later. Keeping track with the DONE system should really be a no brainer.
1128 But there is something more: the whole filesystem is a database and the
1129 recentfiles are one possible representation of it. It's a pretty useful
1130 representation I think that's why I have implemented something around
1131 it. But for strictly local operation it has little value. For local
1132 operation we would much rather have a database. So we would enter every
1133 recentfile reading and every rsync operation and for every file the last
1134 state change and what it leads to. Then we would always ignore older
1135 records without the efforts involved with recentfiles.
1137 The database would have: path,recentepoch,rsyncedon,deletedon
1139 Oh well, not yet clear where this leads to.
1141 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1143 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1144 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1146 * statusfile, maybe only on demand, alone to have a sharp debugging
1147 tool. It is locked and all recentfiles dump themselves into it and we
1148 can build a viewer that lets us know where we stand and what's inside.
1150 * remember: only the principal recentfile needs expiration, all others
1151 shall be expired by principal if it discovers that something has move
1152 upstream.
1154 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1156 * Always check if we stringify to a higher value than in the entry
1157 before. DONE
1159 * And in covered make an additional check if we would be able to see a
1160 numerical difference between the two numbers and if we can't then switch
1161 to a different, more expensive algorithm. Do not want to be caught by
1162 floating surprises. DONE
1164 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1166 * caching has several aspects here: we can cache the interval of the
1167 recentfile which only will change when the mtime of the file changes. We
1168 must re-mirror the recentfile when its ttl has expired. Does have_read
1169 tell you anything? It counts nothing at all. Only the mtime is
1170 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1171 thing the have_mirrored because it is unrelated to the mtime.
1173 * Robustness of floating point calculations! I always thought that the
1174 string calculated by the origin server for the floating representation
1175 of the epoch time is just a string. When we convert it to a number and
1176 later back to a string, the other computer might come to a different
1177 conclusion. This must not happen, we want to preserve it under any
1178 circumstances. I will have to write tests with overlong sequences that
1179 get lost in arithmetic and must see if all still works well. DONE
1181 But one fragile point remains: if one host considers a>b and the other
1182 one considers them == but no eq. To prevent this, we must probably do
1183 some extra homework. DONE
1185 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1187 * the concept of tracking DONE needs an object per recentfile that has
1188 something like these methods:
1190 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1192 covered() register() covered()
1194 The unclear thing is how we translate points in time into intervals. We
1195 could pass a reference to the current recent_events array when running
1196 we_have(xxx) and let the DONE object iterate over it such that it only
1197 has to store a list of intervals that can melt into each other. Ah, even
1198 passing the list together with a list of indexes seems feasiable.
1200 Or maybe ask for the inverted list?
1202 Whenever the complete array is covered by the interval we say we are
1203 fully covered and if the recentfile is not expired, we are uptodate.
1205 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1207 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1209 * need a way to "return" the next entry after the end of a list. When
1210 the caller says "before" or "after" we would like to know if he could
1211 cover that interval/threshold or not because this influences the effect
1212 of a newer timestamp of that recentfile. DONE with $opt{info}.
1214 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1216 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1218 loop implies tixu (?). Nope, something like --statefile decides. Per
1219 default we do ...?
1221 T test, I init, X including removals, U nodirtymark
1223 So we have no concept of dirtymarks, we only trust that since we are
1224 running we have observed everything steadily. But people will not let
1225 this program run forever so we must consider both startup penalty and
1226 book keeping for later runs. We keep this for later. For now we write a
1227 long running mirror that merges several intervals.
1229 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1231 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1232 the intervals!
1234 * rersyncrecent, the script: default to one week. The name of the switch
1235 is --after. Other switches? --loop!
1237 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1239 * need a switch --skip-deletes (?)
1241 * need a switch --enduser that tells us that the whole tempfile
1242 discipline is not needed when there is no downstream user. (?)
1244 Without this switch we cannot have a reasonable recent.pl that just
1245 displays the recent additions. Either we accept to download everything.
1246 Or we download temporary files without the typical rsync protocol
1247 advantages.
1249 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1250 File::Temp::tempdir, this might be a win.
1252 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1254 * apropos missing: we have no push, we never know the downstream
1255 servers. People who know their downstream hosts and want to ascertain
1256 something will want additional methods we have never thought about, like
1257 update or delete a certain file.
1259 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1261 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1262 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1263 difference.
1265 * rsync options: it's a bit of a pain that we usually need several rsync
1266 options, like compress, links, times, checksum and that there is no
1267 reasonable default except the original rsync default. I think wee can
1268 safely assume that the rsync options are shared between all recentfile
1269 instances within one recent tree.
1271 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1273 * deletes: if a delete follows an add quickly enough it may happen that
1274 a downstream mirror did not see the add at all! It seems this needs to
1275 be mentioned somewhere. The point here is that even if the downstream is
1276 never missing the principal timeframe it may encounter a "delete" that
1277 has no complimentary "add" anywhere.
1279 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1281 * I suspect the treat of metadata is incorrect during read or something.
1282 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1283 contained more than 6 hours worth of data. At 06:08 we merged into the
1284 1d file. We need to take snapshots of the 6h file over the course of an
1285 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1286 enough.
1288 Much worse: watching the 1h file: right at the moment (at 06:35) it
1289 covers 1218867584-1219120397 which is 70 hours.
1291 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1292 08:19:44 2008, that is when I checked out last time, so it seems to be
1293 aggregating and never truncating?
1295 No, correct is: it is never truncating; but wrong is: it is aggregating.
1296 It does receive a lot of events from time to time from a larger file.
1297 Somehow a large file gets merged into the small one and because the
1298 "meta/merged" attribute is missing, nobody is paying attention. I
1299 believe that I can fix this by making sure that metadata are honoured
1300 during read. DONE and test adjusted.
1302 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1304 * grand renaming plan
1306 remotebase => remoteroot to fit well with localroot DONE
1307 local_path() => localroot seems to me should already work DONE
1308 recentfile_basename => rfilename no need to stress it has no slash DONE
1310 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1311 kern? It would anyway need a deprecation cycle because it is an
1312 important constructor.
1314 * I like the portability that Data::Serializer brings us but the price
1315 is that some day we might find out that it is slowing us a bit. We'll
1316 see.
1318 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1320 * should we not enter the interval of the principal (or the interval of
1321 the merging file?) in every aggregated/merged file?
1323 * we should aim at a first release and give up on thinking about
1324 sanitizing stuff and zloop. Let's just admit that a full traditional
1325 rsync is the only available sanitizer ATM. Otherwise it's complicated
1326 stuff: sanitizing on the origin server, sanitizing on the slaves,
1327 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1328 get the basics out before this becomes a major cause for mess.
1330 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1332 * On OSes not supporting symlinks we expect that RECENT.recent contains
1333 the contents of the principal recentfile. Actually this is identical on
1334 systems supporting symlinks. Simple, what follows from that is that we
1335 need to keep the serializer in the metadata because we cannot read it
1336 from the filename, doesn't it? Of course not. It's a chicken and egg
1337 problem. This leaves us with the problem to actually parse the
1338 serialized data to find out in which format it is. So who can do the 4
1339 or 5 magics we wanted to support? File::LibMagic?
1341 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1343 * remotebase and recentfile_basename are ugly names. Now that we need a
1344 word for the shortest/principal/driving recentfile too we should do
1345 something about it.
1347 localroot is good. rfile is good. local_path() is bad, local_path($path)
1348 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1349 already deprecated.
1351 Up to now remotebase was the string that described the remote root
1352 directory in rsync notation, like pause.perl.org::authors. And
1353 recentfile_basename was "RECENT-1h.yaml".
1355 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1357 * The test that was added in today's checkin is a good start for a test
1358 of rmirror. We should have more methods in Recent.pm: verify,
1359 addmissingfiles. We should verify the current tree, then rmirror it and
1360 then verifytree the copy. We could then add some arbitrary file and let
1361 it be discovered by addmissingfiles, then rmirror again and then
1362 verifytree the copy again.
1364 Then we could start stealing from csync2 sqlite database [no port to
1365 OSX!] and fill a local DB. And methods to compare the database with the
1366 recentfiles. Our strength is that in principle we could maintain state
1367 with a single float. We have synced up to 1234567890.123456. If the Z
1368 file does not add new files all we have to do is mirror the new ones and
1369 delete the goners.
1371 This makes it clear that we should extend current protocol and declare
1372 that we cheat when we add files too late, just to help the other end
1373 keeping track. Ah yes, that's what was meant when zloop was mentioned
1374 earlier.
1376 Maybe need to revisit File::Mirror to help me with this task.
1378 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1380 * There must be an allow-me-to-truncate flag in every recentfile.
1381 Without it one could construct a sequence of updates winning the locking
1382 battle against the aggregator. Only if an aggregator has managed to
1383 merge data over to the next level, truncating can be allowed. DONE with
1384 accessor merged.
1386 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1388 * We should probably guarantee that no duplicates enter the aggregator
1389 array.
1391 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1393 * To get merge operation faster would need a good benchmark test. What
1394 02 spits out isn't reliable enough and is dominated by many other
1395 things. Between
1397 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1398 Author: Andreas J. Koenig <k@k75.(none)>
1399 Date: Sat Aug 2 07:58:04 2008 +0200
1401 and
1403 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1404 Author: Andreas J. Koenig <k@k75.(none)>
1405 Date: Sat Aug 2 11:40:29 2008 +0200
1407 there was a lot of trying but the effect is hardly measurable with
1408 current tests.
1410 * overhead of connecting seems high. When setting
1411 max_files_per_connection to 1 we see that.
1413 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1415 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1417 25138.2215600014/3600 = 6.98283932222261
1419 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1420 be a bug in the aggregator. FIXED
1422 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1424 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1425 we have it, can we mirror it, mirror it!
1427 I fear this needs a new class which might be called
1428 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1429 and treat them as an entity. I realize that a single recentfile may be
1430 sufficient for certain tasks and that it is handy for the low level
1431 programmer but it is not nice to use. If there is a delete in the 1h
1432 file then the 6h file still contains it. Seekers of the best information
1433 need to combine at least some of the recentfiles most of the time.
1435 There is the place for the Z loop!
1437 But the combination is something to collect in a database, isn't it. Did
1438 csync2 just harrumph?
1440 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1442 * it just occurred to me that hosts in the same mirroring pool could
1443 help out each other even without rewriting the recentfile. Just fetch
1444 the stuff to mirror from several places, bingo. But that's something
1445 that should rather live in a separate package or in rsync directly.
1447 * cronjobs are unsuited because with ntp they would all come at the full
1448 minute and disturb each other. Besides that I'd hate to have a backbone
1449 with more than a few seconds latency.
1451 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1453 * a second rsync server with access control for PAUSE. Port? 873 is the
1454 standard port, let's take 8873.
1456 * if there were a filesystem based on this, it would have a slow access
1457 to inexistent files. It would probably provide wrong readdir (only based
1458 on current content) or also a slow one (based on a recentfile written
1459 after the call). But it would provide fast access to existing files. Or
1460 one would deliberately allow slightly blurred answers based on some
1461 sqlite reflection of the recentfiles.
1463 * todo: write a variant of mirror() that combines two or more
1464 recentfiles and treats them like one
1466 * todo: signal handler to remove the tempfile
1468 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1470 * now that we have the symlink I forgot how it should be used in
1471 practice.
1473 * the z loop: add missing files to Z file. Just append them (instead of
1474 prepending). So one guy prepends something from the Y file from time to
1475 time and another guy appends something rather frequently. Collecting
1476 pond. When Y merges into Z, things get epoch and the collecting pond
1477 gets smaller. What exactly are "missing files"?
1479 take note of current epoch of the alpha file, let's call it the
1480 recent-ts
1482 find all files on disk
1484 remove all files registered in the recentworld up to recent-ts
1486 remove all files that have been deleted after recent-ts according to
1487 recentworld
1489 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1491 * rersyncrecent might be a cronjob with a (locked) state file which
1492 contains things like after and maybe last z sync or such?
1494 rrr-mirror might be an alternative name but how would we justify the
1495 three Rs when there is no Re-Rsync-Recent?
1497 With the --loop parameter it is an endless loop, without it is no loop.
1498 At least this is simple.
1500 * todo: new accssor z-interval specifies how often the Z file is updated
1501 against the filesystem. We probably want no epoch stamp on these
1502 entries. And we want to be able to filter the entries (e.g. no
1503 by-modules and by-category tree)
1505 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1507 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1508 into Z. We do not need the other files filled up with filesystem
1509 contents though.
1511 * need interface to query for a file in order to NOT call update on
1512 PAUSE a second time within a short time.
1514 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1516 * recommended update interval? Makes no sense, is different for
1517 different users.
1519 * Moosify
1521 Local Variables:
1522 mode: change-log
1523 change-log-default-name: "Todo"
1524 tab-width: 2
1525 left-margin: 2
1526 End: