1 2011-02-21 Andreas J. Koenig <andk@cpan.org>
3 * rrr-server has a memory leak
5 root 16063 94.7 10.6 358404 354600 pts/21 RN Feb18 4117:16 /home/src/perl/repoperls/installed-perls/perl/v5.13.8-16-gf53580f/bin/perl -Ilib bin/rrr-server --verbose /home/ftp/incoming/RECENT-1h.yaml
7 experimenting with a fork for aggregate():
9 root 12196 60.2 0.2 13652 9920 pts/21 RN 08:25 1:56 /home/src/perl/repoperls/installed-perls/perl/v5.13.8-16-gf53580f/bin/perl -Ilib bin/rrr-server --verbose /home/ftp/incoming/RECENT-1h.yaml
11 A delete of several hundred files:
13 root 12196 72.8 0.3 13792 10056 pts/21 RN 08:25 3:48 /home/src/perl/repoperls/installed-perls/perl/v5.13.8-16-gf53580f/bin/perl -Ilib bin/rrr-server --verbose /home/ftp/incoming/RECENT-1h.yaml
15 Delete several hundred again:
17 root 12196 79.8 0.3 14788 10944 pts/21 RN 08:25 6:34 /home/src/perl/repoperls/installed-perls/perl/v5.13.8-16-gf53580f/bin/perl -Ilib bin/rrr-server --verbose /home/ftp/incoming/RECENT-1h.yaml
21 2011-02-20 Andreas J. Koenig <andk@cpan.org>
23 * start consistently considering the RECENT files themselves not part of
26 2011-02-19 Andreas J. Koenig <andk@cpan.org>
28 * Found the term anonymous one way file system
30 * No support for empty directories
32 2011-02-17 Andreas J. Koenig <andk@cpan.org>
34 * .recent/ or .rrr/ for metadata
36 * ALERT rrr: we have a piece of code somewhere in the mirror()
37 subroutine that reads the raw YAML file with normal open() and cuts it
38 off for efficiency, then feeds it to yaml::load in order to get some
39 metadata. Reading the thing with the whole tail of the recent array
40 kills the whole idea. The whole thing will not work with JSON.
42 So we must invent a protocol 2 that works with one additional index
43 files, say "A", only containing meta.
45 2011-02-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
47 * I think interval guaranteed with the $last_aggregate_call
48 variable (both in fsck and server) need to be fixed with a $recc
49 variable to something like 60 seconds or even less.
51 * minor bug: fsck just added the lockfile to the index which should be
52 considered bookkeeping.
54 * speed: a large directory remove that is done by the kernel in 1
55 seconds triggers minutes of bookkeeping (07:56:42 - 08:01:10 for ~3000
56 files, all due to Schlemiehl again. Want to do some collecting of
57 pending ops before actually writing to the RECENT files. Lock!
59 Repeating the timing with 2600 files and it took 21:28:08 - 21:35:27.
61 Now having rewritten the loop to use batch_update(): 4383 file removed
62 21:43:37 - 21:44:12. From 7:30 to 0:35 while doing 60% more work that's
65 * lockdirectory expiration? server died and blocked fsck for so long.
67 * At the moment rrr-init is not needed anymore, rrr-server can be
68 started and followed by rrr-fsck. Still stupid but not the main
71 2011-02-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
73 * IN_Q_OVERFLOW needs to trigger an fsck, likewise the entry into the
76 2011-01-30 Andreas J. Koenig <andk@cpan.org>
78 * rrr-init and rrr-server need to be used such:
84 Stupid but no showstopper ATM.
86 2010-10-26 Andreas J. Koenig <andk@cpan.org>
88 * forking slave is done, needs more testing.
90 2010-07-06 Andreas J. Koenig <andk@cpan.org>
92 * The two next most important goals are to get a useable rrr-server and
93 a forking slave implementation that uses virtually no memory after the
94 first pass through. Does it need the status file?
96 * rrr-init now being fast we can now remove the alpha alerts. Done.
98 2010-06-16 Andreas J. Koenig <andk@cpan.org>
100 * rrr-init seems to be sssslllloooowwww. At 3:49 I started it for the
101 video tree. After twelve minutes the file has reached a size of 500k. It
102 is locking and rewriting the YAML after every file. Wasn't there an
103 unlocked variant? For "init" one would not expect locked correctness and
104 a rewrite after every item. Given that this tends to become slower with
105 every file it is really unacceptable.
107 Reminds me of rrr-dirtyupdate (the former name of rrr-update). It seems
108 the code was not dirty enough and that was the reason why it was so
109 slow. And rrr-update does not have a --dirty switch, so it looks like we
110 have no way to force dirty and fast operation yet but we need it for
113 Irritating the bootstrapping race: the order {(1) fill initial
114 array, (2) start server} leaves the need for an fsck to cover the time
115 between 1 and 2. Given that regular fsck is needed in any case the price
116 is not high but still doesn't give a warm feeling. Maybe starting the
117 server will by itself do an fsck?
119 2010-06-15 Andreas J. Koenig <andk@cpan.org>
121 * remove the wild alpha alerts and say, it is alpha but parts of it are
122 already working very well or so.
124 2009-12-10 Andreas J. Koenig <andk@cpan.org>
126 * Prominent request for statusfile to safe memory consumption. I'd like
127 to work backwards: get a verification program that verifies how well in
128 sync two nodes are without using frmr itself. Pretend that the forking
129 agent already works and can be interrupted with ^C. Then let it die at
130 random and restart at random and see the memory consumption and the
131 failure rate over time. Compare it with the curve we have now.
133 Of course we also invented the status file as a sharp debugging tool.
134 I'd like to see it in action. See also: F:R:M:Recent::thaw().
136 2009-06-22 Andreas J. Koenig <andk@cpan.org>
138 * Todo: rrr-init to generate some index files so that rrr-server can be
141 2009-05-02 Andreas J. Koenig <andk@cpan.org>
143 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
144 so we can use it for normal update with a single file. And add a cronjob
145 that runs rrr-fsck so we get a reminder when something is not working
148 2009-04-27 Andreas J. Koenig <andk@cpan.org>
150 * Todo: I still see more link_stat errors than expected. I come back to
151 the former suspicion that we do not do necessary bookkeeping on deleted
152 files so that when we reach the "new" event on a meanwhile deleted file
153 we try to mirror it although we could know better. It's just noise, no
154 harm but noise is irritating.
156 * Todo: more experimenting with runstatusfile to get it official
158 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
161 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
163 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
165 And that ignore_link_stat_error is propagated in sparse_clone.
169 For the record: the bug was that ignore_link_stat_error was lost during
170 sparse_clone. And then the second bug was that it did not default to
171 true. Both are now fixed.
173 2009-04-25 Andreas J. Koenig <andk@cpan.org>
175 * Todo: fill rrr program with life
177 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
179 * Minor bug: I see complaints about files not existing like
180 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
181 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
183 In the RECENT files these are delete events but of course also new
184 events in older recent files.
186 I think we do not properly remember deletes when we skip-deletes and so
187 do not filter them out when we later see the "new" event.
189 Yes, but that what ignore_link_stat_error is for, we know about this
190 race condition. We must make ignore_link_stat_error default to true.
194 * bug in Done.pm failing to merge two fields into one when a third is
195 present. FIXED and accompanied with new test.
197 2009-04-24 Andreas J. Koenig <andk@cpan.org>
199 * Todo: keep the index clean and run some kind of nondestructive fsck 4
202 * who is our backbone?
204 http://cpan.cpantesters.org/authors/02STAMP # barbie
205 http://cpan.solfo.com/authors/02STAMP # abh
206 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
208 2009-04-17 Andreas J. Koenig <andk@cpan.org>
210 * the changelog helper I included in the Makefile release memo does not
211 show tags and is much less useful than I thought. gitk is probably much
214 2009-04-16 Andreas J. Koenig <andk@cpan.org>
218 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
219 #### Ydr_.yaml ... DONE
220 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
221 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
222 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
223 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
224 #### gone already at cpan-pause.pl line 0
226 Running without skip-deletes now but cannot reproduce.
228 2009-04-15 Andreas J. Koenig <andk@cpan.org>
230 * consider state directory such we can restart after a ^C or come again
233 Reuse _runstatusfile: write atomically, see how we can use the dump to
234 start running again. Do not leak and remove globs before writing. Do not
235 only dump one rf, dump the whole complex.
237 * consider the effect of ^C. Is it safe?
239 * want to use Perl::Version to up the versions on every release. Only if
240 things have changed, of course. Does this actually work?
242 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
243 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
245 Must do the bookkeeping myself when this should happen, maybe based on git?
247 I think I prefer the same version number for all pm files. But one also
248 needs a routine for setting them. Ahh:
250 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
256 cpan[2]> m /Mirror::Recent/
257 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
258 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
259 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
260 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
263 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
265 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
267 I just tried MM 6.5102 and it again produces only a provides for Recent,
268 not for the other modules.
270 My own fault. FIXED in the Makefile.PL.
272 2009-04-13 Andreas J. Koenig <andk@cpan.org>
274 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
276 * is it true (as stated at
277 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
278 that rsync can lead to errors when upstream changes before client sync
279 has completed? Can we see the error when we run this on upstream:
282 use Time::HiRes qw(time);
284 open my $fh, ">", "changingfile.txt.new" or die;
285 print $fh (time."\n") x 1000000;
287 rename "changingfile.txt.new", "changingfile.txt" or die;
291 And on the receiving end:
294 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
298 I cannot get it to fail. Apparently it is sufficient when upstream
299 always writes atomically (which of course is mandatory)?
301 SENT to mailing list Apr 28.
305 > https://fedorahosted.org/InstantMirror/
306 > https://www.redhat.com/mailman/listinfo/instantmirror-list
307 > irc.freenode.net channel #instantmirror
309 A cool name for a project. Inspires me to write a few sentences about
310 bittorrent's role in the grand picture.
312 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
313 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
315 2009-04-12 Andreas J. Koenig <andk@cpan.org>
317 * Interesting last minute bug during real download testing: the output
318 isn't as pretty anymore, it seems that more work is being done than
321 Introducing a highlevel _alluptodate method helped a lot in debugging
324 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
325 different on 1h file than on the other files. This lead to frequent
326 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
329 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
331 * Todo: try if we can be faster with a native float package. I'm really
332 glad to have a machine exposing perl floating point bugs, it forced me
333 to retract this item from the changes file: "several speedups in the
334 fakebigfloat code" before I get cpan testers fails.
336 2009-04-11 Andreas J. Koenig <andk@cpan.org>
338 * Bug: when the dirtymark changes upstream, then the downstream server
339 notices it immediately and mirrors the "1h" file but then leaves the
340 rmirror loop and continues with "6h" on the next round through. This
341 behaviour goes on until it reaches the Z file. Because we eat the Z file
342 piecemeal we leave the loop after a while but after that everybody seems
343 to have forgotten that there is some work left to be done. Bad, bad,
346 ETOOCONFUSING better name?
348 Recentfile::get_remote_recentfile_as_tempfile OK
349 Recentfile::resolve_recentfilename split_rfilename DONE
351 Recent::_principal_recentfile_object _principal_recentfile_fromremote
352 Recent::principal_recentfile OK
353 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
355 some void, some not, some not void but called in void context.
357 2009-04-10 Andreas J. Koenig <andk@cpan.org>
359 * Bug? When the dirtymark changes then the second tier hosts quickly
360 reset their DONE state and restart mirroring. But then they mirror
361 potentially outdated index files with inconsistent dirtymark because the
362 first tier box may be a bit behind on the large recentfiles.
364 This means we need a brake when iterating over the recentfiles that
365 refuses to use a recentfile with the wrong dirtymark.
367 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
369 * install the bin/* files and talk about rrr-overview somewhere.
371 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
372 created outside the target tree.
374 * Barbie asks for better logging capabilities
376 2009-03-29 Andreas J. Koenig <andk@cpan.org>
378 * known bug: cannot configure the Recentfile objects to keep delete
379 events. How to call the accessor for that? keep_delete_objects_forever?
381 Sounds acceptable. FIXED (but untested).
383 * known bug: keeps temporary index files lying around.
385 2009-03-24 Andreas J. Koenig <andk@cpan.org>
387 * Some equivalent for
389 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
393 2009-03-22 Andreas J. Koenig <andk@cpan.org>
395 * k81 is again client of k75 with different parameters.
397 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
398 about it all the time and nobody reacts accordingly; (2) lots of temp
399 files get created and not removed; culprit the new call to
400 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
401 caller has to remove the tempfile after use.
403 Need the drawing board.
405 PARTLY FIXED: retracting the idea to call
406 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
407 and unseed such that uptodate now means "we have mirrored this
408 recentfile's bunch of files" and seeded means "somebody believes the
409 index file for this rf needs be refreshed". This seems to work now.
411 Still collecting temp files that nobody cares to depollute.
413 2009-03-21 Andreas J. Koenig <andk@cpan.org>
415 * Bug?: should it be harder than it is atm to set the timestamp to the
418 * bug with native integers:
422 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
426 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
429 Native integer broke when native math was turned off. FIXED
431 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
432 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
435 Yes, records out of order:
438 447004 epoch: 995885533
439 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
442 447008 epoch: 995890358
443 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
446 447012 epoch: 995892221
447 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
450 FIXED with sanity check and later with the integer fix.
452 * Bug: want the index files in a .recent directory
454 * Bug: lots of dot files are not deleted in time
456 * possible test case: can a delete change the timestamp? This would
457 probably break the order of events.
459 2009-03-20 Andreas J. Koenig <andk@cpan.org>
461 * 1233701831.34486 what's so special about this number/string? It
463 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
464 atm lives in Y,Q, and Z.
466 It is the first entry after id/--skip-locking which has timestamp
467 1234164228.11325 which represents a file that doesn't exist anymore.
471 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
472 _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
473 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
474 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
475 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
476 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
477 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
478 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
479 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
480 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
481 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
482 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
483 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
485 FIXED, it was the "--skip-locking" file where manual intervention was participating
487 * bug on the mirroring slave: when the dirtymark gets increased we
488 probably do not reset the done intervals. The mirrorer stays within
489 tight bounds where it tries to sync with upstream and never seems to
490 finish. In the debugging state file I see lots of identical intervals
491 that do not get collapsed. When I restart the mirrorer it dies with:
493 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
494 _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
495 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
496 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
497 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
498 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
499 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
500 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
501 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
502 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
504 and debugging stands at
510 epoch: 1237400802.5789
511 path: id/J/JO/JOHND/CHECKSUMS
514 epoch: 1237400807.97514
515 path: id/M/MI/MIYAGAWA/CHECKSUMS
518 epoch: 1237400817.94363
519 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
527 and it is reproducable.
529 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
530 pause has a fresh one.
532 FIXED, it was the third anded term in each of the ifs in the IV block in
533 _register_one: with that we make sure that we do not stamp on valuable
536 2009-03-17 Andreas J. Koenig <andk@cpan.org>
538 * done: verified the existence of the floating point bug in bleadperl
539 and verified that switching from YAML::Syck to YAML::XS does not resolve
542 BTW, the switch was doable with
544 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
546 and should be considered as a separate TODO
548 * todo: integrate a dirty update with two aggregate calls before
549 unlocking for frictionless dirtying DONE
551 * todo: start the second rsync daemon on pause DONE
553 * todo: move index files to .recent: this cannot simply be done by
554 setting filenameroot to .recent/RECENT. Other parts of the modules rely
555 on the fact that dirname(recentfile) is the root of the mirrored tree.
557 2009-03-16 Andreas J. Koenig <andk@cpan.org>
559 * What was the resolution of the mirror.pl delete hook bug? Do we call
560 the delete hook when pause removes a file from MUIR?
562 * Today on pause: Updating 2a13fba..29f284d and installing it for
563 /usr/local/perl-5.10.0{,-RC2}
565 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
566 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
568 2009-03-15 Andreas J. Koenig <andk@cpan.org>
570 * currently recent_events has the side effect of setting dirtymark
571 because it forces a read on the file. That should be transparent, so
572 that the dirtymark call always forces a cache-able(?) read.
574 * The bug below is -- after a lot of trying -- not reproducible on a
575 small script, only in the large test script. The closest to the output
581 my $x = "01237123229.8814";
586 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
588 $l = "1237123231.22458";
589 $r = "1237123231.22458";
591 Devel::Peek::Dump $l;
592 Devel::Peek::Dump $r;
593 Devel::Peek::Dump $x = $l <=> $r;
597 The checked in state at c404a85 fails the test with my
598 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
599 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
600 in the test script always consistent.
602 * Todo: write a test that inserts a second dirty file with an already
603 existing timestamp. DONE
605 * Bug in perl 5.10 on my 64bit box:
607 DB<98> Devel::Peek::Dump $l
608 SV = PVMG(0x19e0450) at 0x142a550
610 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
612 NV = 1237123231.22458
613 PV = 0x194ce70 "1237123231.22458"\0
617 DB<99> Devel::Peek::Dump $r
618 SV = PVMG(0x19e0240) at 0x142a3e8
620 FLAGS = (PADMY,POK,pPOK)
623 PV = 0x19ff900 "1237123231.22458"\0
627 DB<100> Devel::Peek::Dump $l <=> $r
628 SV = IV(0x19ea6e8) at 0x19ea6f0
630 FLAGS = (PADTMP,IOK,pIOK)
633 DB<101> Devel::Peek::Dump $l
634 SV = PVMG(0x19e0450) at 0x142a550
636 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
638 NV = 1237123231.22458
639 PV = 0x194ce70 "1237123231.22458"\0
643 DB<102> Devel::Peek::Dump $r
644 SV = PVMG(0x19e0240) at 0x142a3e8
646 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
648 NV = 1237123231.22458
649 PV = 0x19ff900 "1237123231.22458"\0
653 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
655 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
657 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
659 * $obj->merge ($other) needs to learn about equal epoch which may happen
660 since dirty_epoch intruded.
662 * Wontfix anytime soon: I think we currently do not support mkdir. Only
665 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
667 * Todo: continue working on update(...,$dirty_epoch). It must be
668 followed by a fast_aggregate! DONE
670 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
672 * maybe we need a closest_entry or fitting_interval or something like
673 that. We want to merge an event into the middle of some recentfile.
674 First we do not know which file, then we do not know where to lock,
675 where to enter the new item, when and where to correct the dirtymark.
677 So my thought is we should first find which file.
679 Another part of my brain answers: what would happen if we would enter
680 the new file into the smallest file just like an ordinary new event,
681 just as an old event?
683 (1) we would write a duplicate timestamp? No, this would be easy to
686 (2) we would make the file large quickly? Yes, but so what? We are
687 changing the dirtymark, so are willing to disturb the downstream hosts.
689 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
691 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
695 https://rt.cpan.org/Ticket/Display.html?id=41199
697 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
699 Strangest output being HAKANARDO who managed to upload
701 Here is a better oneliner that includes also the first line of each
704 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
706 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
708 * memory leak in the syncher? It currently weighs 100M.
712 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
715 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
717 * lookup by epoch and by path and use this ability on the pause to never
718 again register a file twice that doesn't need it. Let's call it
721 * after the dirtymark is done: fill up recentfiles with fake (historic)
722 entries; fill up with individual corrections; algorithm maybe to be done
723 with bigfloat so that we can always place something in the middle
724 between two entries. Before we must switch to bigfloat we could try to
725 use Data::Float::nextup to get the.
727 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
730 * dirtymark now lives in Recentfile, needs to be used in rmirror.
732 * find out why the downloader died after a couple of hours without a net
733 connection. Write a test that survives the not-existence of the other
736 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
738 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
739 it makes less sense with tight coupling for secondary files. But asking
740 the server what the current epoch is might be cheaper on HTTP than on
741 rsync. (Needs to be evaluated)
743 * remove the 0.00 from the verbose overview in the Merged column in the
746 * write tests that expose the problems of the last few days: cascading
747 client/server roles, tight coupling for secondary RFs, deletes after
750 * Some day we might want to have policy options for the slave:
751 tight/loose/no coupling with upstream for secondary RFs. tight is what
752 we have now. loose would wait until a gap occurs that can be closed.
754 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
756 * revisit all $rfs->[$i+1] places if they now make sense still
758 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
760 * another bug is the fact that the mirror command deletes files before
761 it unhides the index file, thus confusing downstream slaves. We must not
762 delete before unhiding and must delete after unhiding. FIXED.
764 * new complication about the slave that is playing a server role.
765 Currently we mirror from newest to oldest with a hidden temporary file
766 as index. And when one file is finished, we unhide the index file.
767 Imagine the cascading server/slave is dead for a day. It then starts
768 mirroring again with the freshest thing and unhides the freshest index
769 file when it has worked through it. In that moment it exposes a time
770 hole. Because it now works on the second recentfile which is still
773 We currently do nothing special to converge after such a drop out. At
774 least not intentionally and robustly and thought through.
776 The algorithm we use to seed the next file needs quite a lot of more
777 robustness than it currently has. Something to do with looking at the
778 merged element of the next rf and when it has dropped off, we seed
779 immediately. And if it ramains dropped off, we seed again, of course.
781 Nope, looking from smaller to larger RFS we look at the merged element
782 of this RF and at the minmax/max element of the next RF. If that
783 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
786 And we need a public accessor seed and unseed or seeded. But not the mix
787 of public and private stuff that then is used behind the back.
789 And then the secondary* stuff must go.
791 And we must understand what the impact is on the DONE system. Can it go
792 unnoticed that there was a hole? And could the DONE system have decided
793 the hole is covered? This should be testable with three directories where
794 the middle stops working for a while. Done->merge is suspicious, we must
795 stop it from merging non-conflatable neighbors due to broken continuity.
799 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
801 * Slaven suggests to have the current epoch or the whole current
802 recentfile available from the HTTP server and take it away with
803 keepalive. This direction goes the granularity down to subseconds.
805 We might want to rewrite everything to factor out transport and allow
806 the whole thing to run via HTTP.
808 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
810 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
813 * smoker on k81 fetching from k75 to verify cascading works. See
814 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
816 * maybe the loop should wait for CHECKSUMS file after every upload. And
817 CPAN.pm needs to deal with timestamps in the future.
819 * do not forget the dirtymark!
821 Text: have a new flag on recentfiles with the meaning: if this
822 changes, you're required to run a full rsync over all the files. The
823 reason why we set it would probably be: some foul happened. we injected
824 files in arbitrary places or didn't inject them although they changed.
825 The content of the flag? Timestamp? The relation between the
826 recentfiles would have to be inheritance from the principal, because any
827 out of band changes would soon later propagate to the next recentfile.
829 By upping the flag often one can easily ruin the slaves.
831 last out of band change? dirtymark?
833 Anyway, this implies that we read a potentially existing recentfile
836 And it implies that we have an eventloop that keeps us busy in 2-3
837 cycles, one for current stuff (tight loop) and one for the recentfiles
838 (cascade when principal has changed), one for the old stuff after a
841 And it implies that the out-of-band change in any of the recentfiles
842 must have a lock on the principal file and there is the place to set the
845 * start a FAQ, especially quick start guide questions. Also to aid those
846 problematic areas where we have no good solution, like the "links"
849 * wish feedback when we are slow.
853 * Remove a few DEBUG statements.
855 * The multiple-rrr way of doing things needs a new option to rmirror,
856 like piecemeal or so. Not urgent because after the first pass through,
857 things run smoothely. It's only ugly during the first pass.
859 * I have the suspicion that the code is broken that decides if the
860 neighboring RF needs to be seeded. I fear when too much time has gone
861 between two calls (in our case more than one hour), it would not seed
862 the neighbor. Of course this will never be noticed, so we need a good
865 * local/localroot confusion: I currently pass both options but one must
868 * accounts for early birds on PAUSE rsync daemon.
870 * hardcoded 20 seconds
872 * who mirrors the index? DOING now.
874 * which CPAN mirrors offer rsync?
876 * visit all XXX, visit all _float places
878 * rename the pathdb stuff, it's too confusing. No idea how.
880 * rrr-inotify, backpan, rrr-register
882 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
884 * current bugs: the pathdb seems to get no reset, the seeding of the
885 secondaryttl stuff seems not to have an effect. Have helped myself with
886 a rand(10), need to fix this back. So not checked in. Does the rand
889 The rand thing helps. The secondaryttl stuff was in the wrong line,
892 The pathdb stuff was because I called either _pathdb or __pathdb on the
893 wrong object. FIXED now.
895 * It's not so beautiful if we never fetch the recentfiles that are not
896 the principal, even if this is correct behaviour. We really do not need
897 them after we have fetched the whole content.
899 OK, we want a switch for that: secondaryttl DONE
901 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
903 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
904 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
905 it lets 40 follow. FIXED
907 * See that the long running process really only updates the principal
908 file unless it has missed a timespan during which something happened. If
909 nothing happened, it must notice even when it misses the timespan. DONE
911 * we must throw away the pathdb when we have reached the end of Z. From
912 that moment we can have a very small pathdb because the only reason for
913 a pathdb is that we know to ignore old records in old files. We won't
914 need this pathdb again before the next full pass over the data is
915 necessary and then we will rebuild it as we go along. DONE
917 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
919 * I think, Done::register_one is doing wrong in that it does not
920 conflate neighboring pieces. The covered() method cannot do this because
921 it has no recent_events array at hand. But register_one has it and could
922 do it and for some reason misses to do it (sometimes).
924 This means that the three tests I just wrote can probably not survive
925 because they test with an already broken Done structure.
927 The art now is to detect how it happens, then to reproduce, then write a
930 So from the logfile this is what happens: we have a good interval with
931 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
932 top of it. Locally we now mirror F2 and open a new done interval for it.
933 Then we mirror F1 but this time with the timestamp T1b. And when we then
934 try to close the gap, we do not find T1 but instead something older. We
935 should gladly accept this older piece and this would fix this bug.
939 * bug to fix: when the 1h file changes while rmirror is running, we do
940 correctly sync the new files but never switch to the 6h file but rather
941 stay in a rather quick loop that fetches the 1h file again and again.
943 Is it possible that we initialize a new object? Or does
944 get_remote_recentfile_as_tempfile overwrite something in myself?
946 Want a new option: _runstatusfile => $file which frequently dumps the
947 state of all recentfiles to a file.
951 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
953 * Todo: now teach update to verify the timestamp is about to write
954 against the previous and use _increase_a_bit if it doesn't comply with
955 strict monotony. DONE
957 * The problem of rounding. So far perl's default precision was
958 sufficient. One day it won't be. FakeFloat has an easy job when it is
959 only reading and other machines have written correctly. But when we want
960 to write a floating point number that is a bit larger than the other
961 one, then we need our own idea of precision.
963 Slaven said: just append a "1". This might be going towards the end of
964 usability too quickly. I'd like something that actually uses the decimal
965 system. Well, appending a 1 also does this but...
967 E.g. we have 1.0. nextup on this architecture is starting with
968 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
969 taken the 1.0000000000000003 and the next user comes and the time tells
970 him 1.0 again. He has to beat my number without stepping over the
971 nextup. This is much less space than I had when I chose 1,2,3.
973 What is also irritating is that nextup is architecture dependent. The
974 128 bit guy must choose very long numbers to fit in between whereas the
975 other one with 16 bit uses larger steps. But then the algorithm is the
976 same for both, so that would be a nice thing.
978 I see two situation where we need this. One is when Time::HiRes returns
979 us a value that is <= the last entry in our recentfile. In this case
980 (let's call it the end-case) we must fill the region between that number
981 and the next higher native floating point number. The other is when we
982 inject an old file into an old recentfile (we would then also set a new
983 dirtymark). We find the integer value already taken and need a slightly
984 different one (let's call it the middle-case). The difference between
985 the two situations is that the next user will want to find something
986 higher than my number in the end-case and something lower than my number
989 So I suggest we give the function both a value and an upper bound and it
990 calculates us a primitive middle. The upper bound in the middle-case is
991 the next integer. The upper bound on the end-case is the nextup floating
992 point number. But the latter poses another problem: if we have occupied
993 the middle m between x and nextup(x), then the nextup(m) will probably
994 not be the same as nextup(x) because some rounding will take place
995 before the nextup is calculated and when the rounding reaches the
996 nextup(x), we will end up at nextup(nextup(x)).
998 So we really need to consider the nextup and the nextdown from there and
999 then the middle and that's the number we may approach asymptotically.
1002 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1004 * consider deprecating the use of RECENT.recent as a symlink. It turns
1005 out to need extra hoops with the rsync options and just isn't worth it.
1006 Or maybe these extra hoops are needed anyway for the rest of the tree?
1007 Nope, can't be the case because not all filesystems support symlinks.
1009 But before doing the large step, I'll deprecate the call of
1010 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
1011 only introduced to resolve RECENT.recent and complicates the routine far
1012 beyond what it deserves.
1014 DONE. Won't deprecate RECENT.recent, just moved its handling to the
1017 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1019 * I think it's a bug that the rsync_option links must be set to true in
1020 order to support RECENT.recent and that nobody cares to set it
1021 automatically. Similar for ignore_link_stat_errors.
1023 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1025 * Todo: find all todos together and make a plan what is missing for a
1028 - verifytree or something like that. fsck maybe.
1030 - rersyncrecent, the script itself? What it do?
1032 - a way to only mirror the recentfiles without mirroring the whole
1033 remote system such that people can decide to mirror only partially see
1034 also 2008-08-30. .shadow-xxx directory? this also needed for a
1035 filesystem that is still incomplete and might need the mirrorfiles for
1038 - long living objects that mirror again and again. Inject something
1039 into ta, see how it goes over to tb.
1041 - how do we continue filling up the DONE system when we use an object
1042 for the second time? "fully covered" and "uptodate" or new terminology.
1044 - overview called on the wrong file should be understandable
1046 - the meta data field that must change when we fake something up so that
1047 the downstream people know they have to re-fetch everything.
1049 - how tolerant are we against missing files upstream? how do we keep
1050 track? there are legitimate cases where we did read upstream index right
1051 before a file got deleted there and then find that file as new and want
1052 it. There are other cases that are not self healing and must be tracked
1055 - how, exactly, do we have to deal with deletes? With rsync errors?
1057 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
1058 authors) failed: No such file or directory (2)
1060 The file above is a delete in 1h and a new in file 1M and the
1061 delete in the locally running rmirror did not get propagated to the 1M
1062 object. Bug. And the consequence is a standstill.
1064 It seems that a slave that works with a file below the principal needs
1065 to merge things all the way up to get rid of later deletes. Or keep
1066 track of all deletes and skip them later. So we need a trackdeletes.pm
1067 similar to the done.pm?
1069 see also 2008-08-20 about spurious deletes that really have no add
1070 counterpart and yet they are not wrong.
1072 - consider the effect when resyncing the recentfile takes longer than
1073 the time per loop. Then we never rsync any file. We need to diagnose
1074 that and force an increase of that loop time. But when we later are fast
1075 enough again because the net has recovered, then we need to switch back
1076 to original parameters. ERm, no, it's enough to keep syncing at least
1077 one file before refetching an index file.
1079 - remember to verify that no temp files are left lying around and the
1082 - status file for not long running jobs that want to track upstream with
1085 - revisit all XXX _float areas and study Sub::Exporter DONE
1087 - persistent DB even though we just said we do not need it. Just for
1088 extended capabilities and time savings when, for example, upstream
1089 announces a reset and we get new recentfiles and could then limit
1090 ourselves to a subset of files (those that have a changed epoch) in a
1091 first pass and would only then do the loop to verify the rest. Or
1094 * Todo: aggregate files should know their feed and finding the principal
1095 should be done stepwise. (?)
1097 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
1098 left locks around. DONE
1100 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1102 * maybe extend the _overview so that it always says if and where the
1103 last file is in the next file and where the next event in the next rf
1104 would lie. No, don't like this anymore. REJECT
1106 * take the two new redundant tests out again, only the third must
1109 * Todo: add a sanity check if the merged structure is really pointing to
1110 a different rf and that this different rf is larger. DONE
1112 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1114 * now test, if they are overlapping. And test if there is a file in the
1115 next rf that would fit into this rf's interval.
1117 1h 1222324012.8474 1222322541.7963 0.4086
1118 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
1119 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
1120 1W 1222313218.3626 1221708477.5829 167.9835
1122 I suspect that somebody writes a merged timestamp without having merged
1123 and then somebody else relies on it.
1125 If aggregate is running, the intervals must not be extravagated, if it
1126 is not running, there must not be bounds, the total number of events in
1127 the system must be counted and must be controlled throughout the tests.
1128 That the test required the additional update was probably nonsense,
1129 because aggregate can cut pieces too. FIXED & DONE
1131 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1133 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1136 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1138 * Most apparent bug at the moment is that the recentfiles are fetched
1139 too often. Only the principal should be fetched and if it has not
1140 changed, the others should not be refetched. ATM I must admit that I'm
1141 happy that we refetch more often than needed because I can more easily
1142 fix bugs while the thing is running.
1144 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1145 done but the done system does not know about it. The reason for the
1146 failure is not known and we never reach the status uptodate because of
1147 this. We must get over it.
1149 Later it turns out that the origin server had a bug somewhere.
1150 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1151 array of the recentfile one position above. The bug was my own.
1153 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1155 * There is the race condition where the server does a delete and the
1156 slave does not yet know and then tries to download it because he sees
1157 the new. So for this time window we must be more tolerant against
1158 failure. If we cannot download a file, we should just skip it and should
1159 not retry immediately. The whole system should discover the lost thing
1160 later. Keeping track with the DONE system should really be a no brainer.
1162 But there is something more: the whole filesystem is a database and the
1163 recentfiles are one possible representation of it. It's a pretty useful
1164 representation I think that's why I have implemented something around
1165 it. But for strictly local operation it has little value. For local
1166 operation we would much rather have a database. So we would enter every
1167 recentfile reading and every rsync operation and for every file the last
1168 state change and what it leads to. Then we would always ignore older
1169 records without the efforts involved with recentfiles.
1171 The database would have: path,recentepoch,rsyncedon,deletedon
1173 Oh well, not yet clear where this leads to.
1175 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1177 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1178 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1180 * statusfile, maybe only on demand, alone to have a sharp debugging
1181 tool. It is locked and all recentfiles dump themselves into it and we
1182 can build a viewer that lets us know where we stand and what's inside.
1184 * remember: only the principal recentfile needs expiration, all others
1185 shall be expired by principal if it discovers that something has move
1188 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1190 * Always check if we stringify to a higher value than in the entry
1193 * And in covered make an additional check if we would be able to see a
1194 numerical difference between the two numbers and if we can't then switch
1195 to a different, more expensive algorithm. Do not want to be caught by
1196 floating surprises. DONE
1198 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1200 * caching has several aspects here: we can cache the interval of the
1201 recentfile which only will change when the mtime of the file changes. We
1202 must re-mirror the recentfile when its ttl has expired. Does have_read
1203 tell you anything? It counts nothing at all. Only the mtime is
1204 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1205 thing the have_mirrored because it is unrelated to the mtime.
1207 * Robustness of floating point calculations! I always thought that the
1208 string calculated by the origin server for the floating representation
1209 of the epoch time is just a string. When we convert it to a number and
1210 later back to a string, the other computer might come to a different
1211 conclusion. This must not happen, we want to preserve it under any
1212 circumstances. I will have to write tests with overlong sequences that
1213 get lost in arithmetic and must see if all still works well. DONE
1215 But one fragile point remains: if one host considers a>b and the other
1216 one considers them == but no eq. To prevent this, we must probably do
1217 some extra homework. DONE
1219 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1221 * the concept of tracking DONE needs an object per recentfile that has
1222 something like these methods:
1224 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1226 covered() register() covered()
1228 The unclear thing is how we translate points in time into intervals. We
1229 could pass a reference to the current recent_events array when running
1230 we_have(xxx) and let the DONE object iterate over it such that it only
1231 has to store a list of intervals that can melt into each other. Ah, even
1232 passing the list together with a list of indexes seems feasiable.
1234 Or maybe ask for the inverted list?
1236 Whenever the complete array is covered by the interval we say we are
1237 fully covered and if the recentfile is not expired, we are uptodate.
1239 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1241 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1243 * need a way to "return" the next entry after the end of a list. When
1244 the caller says "before" or "after" we would like to know if he could
1245 cover that interval/threshold or not because this influences the effect
1246 of a newer timestamp of that recentfile. DONE with $opt{info}.
1248 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1250 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1252 loop implies tixu (?). Nope, something like --statefile decides. Per
1255 T test, I init, X including removals, U nodirtymark
1257 So we have no concept of dirtymarks, we only trust that since we are
1258 running we have observed everything steadily. But people will not let
1259 this program run forever so we must consider both startup penalty and
1260 book keeping for later runs. We keep this for later. For now we write a
1261 long running mirror that merges several intervals.
1263 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1265 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1268 * rersyncrecent, the script: default to one week. The name of the switch
1269 is --after. Other switches? --loop!
1271 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1273 * need a switch --skip-deletes (?)
1275 * need a switch --enduser that tells us that the whole tempfile
1276 discipline is not needed when there is no downstream user. (?)
1278 Without this switch we cannot have a reasonable recent.pl that just
1279 displays the recent additions. Either we accept to download everything.
1280 Or we download temporary files without the typical rsync protocol
1283 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1284 File::Temp::tempdir, this might be a win.
1286 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1288 * apropos missing: we have no push, we never know the downstream
1289 servers. People who know their downstream hosts and want to ascertain
1290 something will want additional methods we have never thought about, like
1291 update or delete a certain file.
1293 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1295 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1296 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1299 * rsync options: it's a bit of a pain that we usually need several rsync
1300 options, like compress, links, times, checksum and that there is no
1301 reasonable default except the original rsync default. I think wee can
1302 safely assume that the rsync options are shared between all recentfile
1303 instances within one recent tree.
1305 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1307 * deletes: if a delete follows an add quickly enough it may happen that
1308 a downstream mirror did not see the add at all! It seems this needs to
1309 be mentioned somewhere. The point here is that even if the downstream is
1310 never missing the principal timeframe it may encounter a "delete" that
1311 has no complimentary "add" anywhere.
1313 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1315 * I suspect the treat of metadata is incorrect during read or something.
1316 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1317 contained more than 6 hours worth of data. At 06:08 we merged into the
1318 1d file. We need to take snapshots of the 6h file over the course of an
1319 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1322 Much worse: watching the 1h file: right at the moment (at 06:35) it
1323 covers 1218867584-1219120397 which is 70 hours.
1325 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1326 08:19:44 2008, that is when I checked out last time, so it seems to be
1327 aggregating and never truncating?
1329 No, correct is: it is never truncating; but wrong is: it is aggregating.
1330 It does receive a lot of events from time to time from a larger file.
1331 Somehow a large file gets merged into the small one and because the
1332 "meta/merged" attribute is missing, nobody is paying attention. I
1333 believe that I can fix this by making sure that metadata are honoured
1334 during read. DONE and test adjusted.
1336 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1338 * grand renaming plan
1340 remotebase => remoteroot to fit well with localroot DONE
1341 local_path() => localroot seems to me should already work DONE
1342 recentfile_basename => rfilename no need to stress it has no slash DONE
1344 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1345 kern? It would anyway need a deprecation cycle because it is an
1346 important constructor.
1348 * I like the portability that Data::Serializer brings us but the price
1349 is that some day we might find out that it is slowing us a bit. We'll
1352 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1354 * should we not enter the interval of the principal (or the interval of
1355 the merging file?) in every aggregated/merged file?
1357 * we should aim at a first release and give up on thinking about
1358 sanitizing stuff and zloop. Let's just admit that a full traditional
1359 rsync is the only available sanitizer ATM. Otherwise it's complicated
1360 stuff: sanitizing on the origin server, sanitizing on the slaves,
1361 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1362 get the basics out before this becomes a major cause for mess.
1364 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1366 * On OSes not supporting symlinks we expect that RECENT.recent contains
1367 the contents of the principal recentfile. Actually this is identical on
1368 systems supporting symlinks. Simple, what follows from that is that we
1369 need to keep the serializer in the metadata because we cannot read it
1370 from the filename, doesn't it? Of course not. It's a chicken and egg
1371 problem. This leaves us with the problem to actually parse the
1372 serialized data to find out in which format it is. So who can do the 4
1373 or 5 magics we wanted to support? File::LibMagic?
1375 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1377 * remotebase and recentfile_basename are ugly names. Now that we need a
1378 word for the shortest/principal/driving recentfile too we should do
1381 localroot is good. rfile is good. local_path() is bad, local_path($path)
1382 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1385 Up to now remotebase was the string that described the remote root
1386 directory in rsync notation, like pause.perl.org::authors. And
1387 recentfile_basename was "RECENT-1h.yaml".
1389 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1391 * The test that was added in today's checkin is a good start for a test
1392 of rmirror. We should have more methods in Recent.pm: verify,
1393 addmissingfiles. We should verify the current tree, then rmirror it and
1394 then verifytree the copy. We could then add some arbitrary file and let
1395 it be discovered by addmissingfiles, then rmirror again and then
1396 verifytree the copy again.
1398 Then we could start stealing from csync2 sqlite database [no port to
1399 OSX!] and fill a local DB. And methods to compare the database with the
1400 recentfiles. Our strength is that in principle we could maintain state
1401 with a single float. We have synced up to 1234567890.123456. If the Z
1402 file does not add new files all we have to do is mirror the new ones and
1405 This makes it clear that we should extend current protocol and declare
1406 that we cheat when we add files too late, just to help the other end
1407 keeping track. Ah yes, that's what was meant when zloop was mentioned
1410 Maybe need to revisit File::Mirror to help me with this task.
1412 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1414 * There must be an allow-me-to-truncate flag in every recentfile.
1415 Without it one could construct a sequence of updates winning the locking
1416 battle against the aggregator. Only if an aggregator has managed to
1417 merge data over to the next level, truncating can be allowed. DONE with
1420 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1422 * We should probably guarantee that no duplicates enter the aggregator
1425 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1427 * To get merge operation faster would need a good benchmark test. What
1428 02 spits out isn't reliable enough and is dominated by many other
1431 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1432 Author: Andreas J. Koenig <k@k75.(none)>
1433 Date: Sat Aug 2 07:58:04 2008 +0200
1437 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1438 Author: Andreas J. Koenig <k@k75.(none)>
1439 Date: Sat Aug 2 11:40:29 2008 +0200
1441 there was a lot of trying but the effect is hardly measurable with
1444 * overhead of connecting seems high. When setting
1445 max_files_per_connection to 1 we see that.
1447 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1449 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1451 25138.2215600014/3600 = 6.98283932222261
1453 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1454 be a bug in the aggregator. FIXED
1456 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1458 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1459 we have it, can we mirror it, mirror it!
1461 I fear this needs a new class which might be called
1462 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1463 and treat them as an entity. I realize that a single recentfile may be
1464 sufficient for certain tasks and that it is handy for the low level
1465 programmer but it is not nice to use. If there is a delete in the 1h
1466 file then the 6h file still contains it. Seekers of the best information
1467 need to combine at least some of the recentfiles most of the time.
1469 There is the place for the Z loop!
1471 But the combination is something to collect in a database, isn't it. Did
1472 csync2 just harrumph?
1474 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1476 * it just occurred to me that hosts in the same mirroring pool could
1477 help out each other even without rewriting the recentfile. Just fetch
1478 the stuff to mirror from several places, bingo. But that's something
1479 that should rather live in a separate package or in rsync directly.
1481 * cronjobs are unsuited because with ntp they would all come at the full
1482 minute and disturb each other. Besides that I'd hate to have a backbone
1483 with more than a few seconds latency.
1485 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1487 * a second rsync server with access control for PAUSE. Port? 873 is the
1488 standard port, let's take 8873.
1490 * if there were a filesystem based on this, it would have a slow access
1491 to inexistent files. It would probably provide wrong readdir (only based
1492 on current content) or also a slow one (based on a recentfile written
1493 after the call). But it would provide fast access to existing files. Or
1494 one would deliberately allow slightly blurred answers based on some
1495 sqlite reflection of the recentfiles.
1497 * todo: write a variant of mirror() that combines two or more
1498 recentfiles and treats them like one
1500 * todo: signal handler to remove the tempfile
1502 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1504 * now that we have the symlink I forgot how it should be used in
1507 * the z loop: add missing files to Z file. Just append them (instead of
1508 prepending). So one guy prepends something from the Y file from time to
1509 time and another guy appends something rather frequently. Collecting
1510 pond. When Y merges into Z, things get epoch and the collecting pond
1511 gets smaller. What exactly are "missing files"?
1513 take note of current epoch of the alpha file, let's call it the
1516 find all files on disk
1518 remove all files registered in the recentworld up to recent-ts
1520 remove all files that have been deleted after recent-ts according to
1523 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1525 * rersyncrecent might be a cronjob with a (locked) state file which
1526 contains things like after and maybe last z sync or such?
1528 rrr-mirror might be an alternative name but how would we justify the
1529 three Rs when there is no Re-Rsync-Recent?
1531 With the --loop parameter it is an endless loop, without it is no loop.
1532 At least this is simple.
1534 * todo: new accssor z-interval specifies how often the Z file is updated
1535 against the filesystem. We probably want no epoch stamp on these
1536 entries. And we want to be able to filter the entries (e.g. no
1537 by-modules and by-category tree)
1539 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1541 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1542 into Z. We do not need the other files filled up with filesystem
1545 * need interface to query for a file in order to NOT call update on
1546 PAUSE a second time within a short time.
1548 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1550 * recommended update interval? Makes no sense, is different for
1557 change-log-default-name: "Todo"