1 2009-12-10 Andreas J. Koenig <andk@cpan.org>
3 * Prominent request for statusfile to safe memory consumption. I'd like
4 to work backwards: get a verification program that verifies how well in
5 sync two nodes are without using frmr itself. Pretend that the forking
6 agent already works and can be interrupted with ^C. Then let it die at
7 random and restart at random and see the memory consumption and the
8 failure rate over time. Compare it with the curve we have now.
10 Of course we also invented the status file as a sharp debugging tool.
11 I'd like to see it in action.
13 2009-06-22 Andreas J. Koenig <andk@cpan.org>
15 * Todo: rrr-init to generate some index files so that rrr-server can be
18 2009-05-02 Andreas J. Koenig <andk@cpan.org>
20 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
21 so we can use it for normal update with a single file. And add a cronjob
22 that runs rrr-fsck so we get a reminder when something is not working
25 2009-04-27 Andreas J. Koenig <andk@cpan.org>
27 * Todo: I still see more link_stat errors than expected. I come back to
28 the former suspicion that we do not do necessary bookkeeping on deleted
29 files so that when we reach the "new" event on a meanwhile deleted file
30 we try to mirror it although we could know better. It's just noise, no
31 harm but noise is irritating.
33 * Todo: more experimenting with runstatusfile to get it official
35 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
38 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
40 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
42 And that ignore_link_stat_error is propagated in sparse_clone.
46 For the record: the bug was that ignore_link_stat_error was lost during
47 sparse_clone. And then the second bug was that it did not default to
48 true. Both are now fixed.
50 2009-04-25 Andreas J. Koenig <andk@cpan.org>
52 * Todo: fill rrr program with life
54 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
56 * Minor bug: I see complaints about files not existing like
57 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
58 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
60 In the RECENT files these are delete events but of course also new
61 events in older recent files.
63 I think we do not properly remember deletes when we skip-deletes and so
64 do not filter them out when we later see the "new" event.
66 Yes, but that what ignore_link_stat_error is for, we know about this
67 race condition. We must make ignore_link_stat_error default to true.
71 * bug in Done.pm failing to merge two fields into one when a third is
72 present. FIXED and accompanied with new test.
74 2009-04-24 Andreas J. Koenig <andk@cpan.org>
76 * Todo: keep the index clean and run some kind of nondestructive fsck 4
79 * who is our backbone?
81 http://cpan.cpantesters.org/authors/02STAMP # barbie
82 http://cpan.solfo.com/authors/02STAMP # abh
83 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
85 2009-04-17 Andreas J. Koenig <andk@cpan.org>
87 * the changelog helper I included in the Makefile release memo does not
88 show tags and is much less useful than I thought. gitk is probably much
91 2009-04-16 Andreas J. Koenig <andk@cpan.org>
95 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
96 #### Ydr_.yaml ... DONE
97 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
98 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
99 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
100 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
101 #### gone already at cpan-pause.pl line 0
103 Running without skip-deletes now but cannot reproduce.
105 2009-04-15 Andreas J. Koenig <andk@cpan.org>
107 * consider state directory such we can restart after a ^C or come again
110 Reuse _runstatusfile: write atomically, see how we can use the dump to
111 start running again. Do not leak and remove globs before writing. Do not
112 only dump one rf, dump the whole complex.
114 * consider the effect of ^C. Is it safe?
116 * want to use Perl::Version to up the versions on every release. Only if
117 things have changed, of course. Does this actually work?
119 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
120 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
122 Must do the bookkeeping myself when this should happen, maybe based on git?
124 I think I prefer the same version number for all pm files. But one also
125 needs a routine for setting them. Ahh:
127 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
133 cpan[2]> m /Mirror::Recent/
134 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
135 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
136 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
137 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
140 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
142 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
144 I just tried MM 6.5102 and it again produces only a provides for Recent,
145 not for the other modules.
147 My own fault. FIXED in the Makefile.PL.
149 2009-04-13 Andreas J. Koenig <andk@cpan.org>
151 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
153 * is it true (as stated at
154 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
155 that rsync can lead to errors when upstream changes before client sync
156 has completed? Can we see the error when we run this on upstream:
159 use Time::HiRes qw(time);
161 open my $fh, ">", "changingfile.txt.new" or die;
162 print $fh (time."\n") x 1000000;
164 rename "changingfile.txt.new", "changingfile.txt" or die;
168 And on the receiving end:
171 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
175 I cannot get it to fail. Apparently it is sufficient when upstream
176 always writes atomically (which of course is mandatory)?
178 SENT to mailing list Apr 28.
182 > https://fedorahosted.org/InstantMirror/
183 > https://www.redhat.com/mailman/listinfo/instantmirror-list
184 > irc.freenode.net channel #instantmirror
186 A cool name for a project. Inspires me to write a few sentences about
187 bittorrent's role in the grand picture.
189 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
190 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
192 2009-04-12 Andreas J. Koenig <andk@cpan.org>
194 * Interesting last minute bug during real download testing: the output
195 isn't as pretty anymore, it seems that more work is being done than
198 Introducing a highlevel _alluptodate method helped a lot in debugging
201 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
202 different on 1h file than on the other files. This lead to frequent
203 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
206 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
208 * Todo: try if we can be faster with a native float package. I'm really
209 glad to have a machine exposing perl floating point bugs, it forced me
210 to retract this item from the changes file: "several speedups in the
211 fakebigfloat code" before I get cpan testers fails.
213 2009-04-11 Andreas J. Koenig <andk@cpan.org>
215 * Bug: when the dirtymark changes upstream, then the downstream server
216 notices it immediately and mirrors the "1h" file but then leaves the
217 rmirror loop and continues with "6h" on the next round through. This
218 behaviour goes on until it reaches the Z file. Because we eat the Z file
219 piecemeal we leave the loop after a while but after that everybody seems
220 to have forgotten that there is some work left to be done. Bad, bad,
223 ETOOCONFUSING better name?
225 Recentfile::get_remote_recentfile_as_tempfile OK
226 Recentfile::resolve_recentfilename split_rfilename DONE
228 Recent::_principal_recentfile_object _principal_recentfile_fromremote
229 Recent::principal_recentfile OK
230 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
232 some void, some not, some not void but called in void context.
234 2009-04-10 Andreas J. Koenig <andk@cpan.org>
236 * Bug? When the dirtymark changes then the second tier hosts quickly
237 reset their DONE state and restart mirroring. But then they mirror
238 potentially outdated index files with inconsistent dirtymark because the
239 first tier box may be a bit behind on the large recentfiles.
241 This means we need a brake when iterating over the recentfiles that
242 refuses to use a recentfile with the wrong dirtymark.
244 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
246 * install the bin/* files and talk about rrr-overview somewhere.
248 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
249 created outside the target tree.
251 * Barbie asks for better logging capabilities
253 2009-03-29 Andreas J. Koenig <andk@cpan.org>
255 * known bug: cannot configure the Recentfile objects to keep delete
256 events. How to call the accessor for that? keep_delete_objects_forever?
258 Sounds acceptable. FIXED (but untested).
260 * known bug: keeps temporary index files lying around.
262 2009-03-24 Andreas J. Koenig <andk@cpan.org>
264 * Some equivalent for
266 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
270 2009-03-22 Andreas J. Koenig <andk@cpan.org>
272 * k81 is again client of k75 with different parameters.
274 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
275 about it all the time and nobody reacts accordingly; (2) lots of temp
276 files get created and not removed; culprit the new call to
277 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
278 caller has to remove the tempfile after use.
280 Need the drawing board.
282 PARTLY FIXED: retracting the idea to call
283 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
284 and unseed such that uptodate now means "we have mirrored this
285 recentfile's bunch of files" and seeded means "somebody believes the
286 index file for this rf needs be refreshed". This seems to work now.
288 Still collecting temp files that nobody cares to depollute.
290 2009-03-21 Andreas J. Koenig <andk@cpan.org>
292 * Bug?: should it be harder than it is atm to set the timestamp to the
295 * bug with native integers:
299 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
303 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
306 Native integer broke when native math was turned off. FIXED
308 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
309 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
312 Yes, records out of order:
315 447004 epoch: 995885533
316 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
319 447008 epoch: 995890358
320 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
323 447012 epoch: 995892221
324 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
327 FIXED with sanity check and later with the integer fix.
329 * Bug: want the index files in a .recent directory
331 * Bug: lots of dot files are not deleted in time
333 * possible test case: can a delete change the timestamp? This would
334 probably break the order of events.
336 2009-03-20 Andreas J. Koenig <andk@cpan.org>
338 * 1233701831.34486 what's so special about this number/string? It
340 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
341 atm lives in Y,Q, and Z.
343 It is the first entry after id/--skip-locking which has timestamp
344 1234164228.11325 which represents a file that doesn't exist anymore.
348 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
349 _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
350 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
351 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
352 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
353 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
354 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
355 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
356 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
357 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
358 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
359 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
360 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
362 FIXED, it was the "--skip-locking" file where manual intervention was participating
364 * bug on the mirroring slave: when the dirtymark gets increased we
365 probably do not reset the done intervals. The mirrorer stays within
366 tight bounds where it tries to sync with upstream and never seems to
367 finish. In the debugging state file I see lots of identical intervals
368 that do not get collapsed. When I restart the mirrorer it dies with:
370 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
371 _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
372 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
373 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
374 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
375 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
376 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
377 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
378 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
379 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
381 and debugging stands at
387 epoch: 1237400802.5789
388 path: id/J/JO/JOHND/CHECKSUMS
391 epoch: 1237400807.97514
392 path: id/M/MI/MIYAGAWA/CHECKSUMS
395 epoch: 1237400817.94363
396 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
404 and it is reproducable.
406 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
407 pause has a fresh one.
409 FIXED, it was the third anded term in each of the ifs in the IV block in
410 _register_one: with that we make sure that we do not stamp on valuable
413 2009-03-17 Andreas J. Koenig <andk@cpan.org>
415 * done: verified the existence of the floating point bug in bleadperl
416 and verified that switching from YAML::Syck to YAML::XS does not resolve
419 BTW, the switch was doable with
421 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
423 and should be considered as a separate TODO
425 * todo: integrate a dirty update with two aggregate calls before
426 unlocking for frictionless dirtying DONE
428 * todo: start the second rsync daemon on pause DONE
430 * todo: move index files to .recent: this cannot simply be done by
431 setting filenameroot to .recent/RECENT. Other parts of the modules rely
432 on the fact that dirname(recentfile) is the root of the mirrored tree.
434 2009-03-16 Andreas J. Koenig <andk@cpan.org>
436 * What was the resolution of the mirror.pl delete hook bug? Do we call
437 the delete hook when pause removes a file from MUIR?
439 * Today on pause: Updating 2a13fba..29f284d and installing it for
440 /usr/local/perl-5.10.0{,-RC2}
442 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
443 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
445 2009-03-15 Andreas J. Koenig <andk@cpan.org>
447 * currently recent_events has the side effect of setting dirtymark
448 because it forces a read on the file. That should be transparent, so
449 that the dirtymark call always forces a cache-able(?) read.
451 * The bug below is -- after a lot of trying -- not reproducible on a
452 small script, only in the large test script. The closest to the output
458 my $x = "01237123229.8814";
463 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
465 $l = "1237123231.22458";
466 $r = "1237123231.22458";
468 Devel::Peek::Dump $l;
469 Devel::Peek::Dump $r;
470 Devel::Peek::Dump $x = $l <=> $r;
474 The checked in state at c404a85 fails the test with my
475 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
476 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
477 in the test script always consistent.
479 * Todo: write a test that inserts a second dirty file with an already
480 existing timestamp. DONE
482 * Bug in perl 5.10 on my 64bit box:
484 DB<98> Devel::Peek::Dump $l
485 SV = PVMG(0x19e0450) at 0x142a550
487 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
489 NV = 1237123231.22458
490 PV = 0x194ce70 "1237123231.22458"\0
494 DB<99> Devel::Peek::Dump $r
495 SV = PVMG(0x19e0240) at 0x142a3e8
497 FLAGS = (PADMY,POK,pPOK)
500 PV = 0x19ff900 "1237123231.22458"\0
504 DB<100> Devel::Peek::Dump $l <=> $r
505 SV = IV(0x19ea6e8) at 0x19ea6f0
507 FLAGS = (PADTMP,IOK,pIOK)
510 DB<101> Devel::Peek::Dump $l
511 SV = PVMG(0x19e0450) at 0x142a550
513 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
515 NV = 1237123231.22458
516 PV = 0x194ce70 "1237123231.22458"\0
520 DB<102> Devel::Peek::Dump $r
521 SV = PVMG(0x19e0240) at 0x142a3e8
523 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
525 NV = 1237123231.22458
526 PV = 0x19ff900 "1237123231.22458"\0
530 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
532 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
534 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
536 * $obj->merge ($other) needs to learn about equal epoch which may happen
537 since dirty_epoch intruded.
539 * Wontfix anytime soon: I think we currently do not support mkdir. Only
542 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
544 * Todo: continue working on update(...,$dirty_epoch). It must be
545 followed by a fast_aggregate! DONE
547 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
549 * maybe we need a closest_entry or fitting_interval or something like
550 that. We want to merge an event into the middle of some recentfile.
551 First we do not know which file, then we do not know where to lock,
552 where to enter the new item, when and where to correct the dirtymark.
554 So my thought is we should first find which file.
556 Another part of my brain answers: what would happen if we would enter
557 the new file into the smallest file just like an ordinary new event,
558 just as an old event?
560 (1) we would write a duplicate timestamp? No, this would be easy to
563 (2) we would make the file large quickly? Yes, but so what? We are
564 changing the dirtymark, so are willing to disturb the downstream hosts.
566 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
568 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
572 https://rt.cpan.org/Ticket/Display.html?id=41199
574 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
576 Strangest output being HAKANARDO who managed to upload
578 Here is a better oneliner that includes also the first line of each
581 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
583 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
585 * memory leak in the syncher? It currently weighs 100M.
589 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
592 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
594 * lookup by epoch and by path and use this ability on the pause to never
595 again register a file twice that doesn't need it. Let's call it
598 * after the dirtymark is done: fill up recentfiles with fake (historic)
599 entries; fill up with individual corrections; algorithm maybe to be done
600 with bigfloat so that we can always place something in the middle
601 between two entries. Before we must switch to bigfloat we could try to
602 use Data::Float::nextup to get the.
604 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
607 * dirtymark now lives in Recentfile, needs to be used in rmirror.
609 * find out why the downloader died after a couple of hours without a net
610 connection. Write a test that survives the not-existence of the other
613 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
615 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
616 it makes less sense with tight coupling for secondary files. But asking
617 the server what the current epoch is might be cheaper on HTTP than on
618 rsync. (Needs to be evaluated)
620 * remove the 0.00 from the verbose overview in the Merged column in the
623 * write tests that expose the problems of the last few days: cascading
624 client/server roles, tight coupling for secondary RFs, deletes after
627 * Some day we might want to have policy options for the slave:
628 tight/loose/no coupling with upstream for secondary RFs. tight is what
629 we have now. loose would wait until a gap occurs that can be closed.
631 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
633 * revisit all $rfs->[$i+1] places if they now make sense still
635 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
637 * another bug is the fact that the mirror command deletes files before
638 it unhides the index file, thus confusing downstream slaves. We must not
639 delete before unhiding and must delete after unhiding. FIXED.
641 * new complication about the slave that is playing a server role.
642 Currently we mirror from newest to oldest with a hidden temporary file
643 as index. And when one file is finished, we unhide the index file.
644 Imagine the cascading server/slave is dead for a day. It then starts
645 mirroring again with the freshest thing and unhides the freshest index
646 file when it has worked through it. In that moment it exposes a time
647 hole. Because it now works on the second recentfile which is still
650 We currently do nothing special to converge after such a drop out. At
651 least not intentionally and robustly and thought through.
653 The algorithm we use to seed the next file needs quite a lot of more
654 robustness than it currently has. Something to do with looking at the
655 merged element of the next rf and when it has dropped off, we seed
656 immediately. And if it ramains dropped off, we seed again, of course.
658 Nope, looking from smaller to larger RFS we look at the merged element
659 of this RF and at the minmax/max element of the next RF. If that
660 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
663 And we need a public accessor seed and unseed or seeded. But not the mix
664 of public and private stuff that then is used behind the back.
666 And then the secondary* stuff must go.
668 And we must understand what the impact is on the DONE system. Can it go
669 unnoticed that there was a hole? And could the DONE system have decided
670 the hole is covered? This should be testable with three directories where
671 the middle stops working for a while. Done->merge is suspicious, we must
672 stop it from merging non-conflatable neighbors due to broken continuity.
676 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
678 * Slaven suggests to have the current epoch or the whole current
679 recentfile available from the HTTP server and take it away with
680 keepalive. This direction goes the granularity down to subseconds.
682 We might want to rewrite everything to factor out transport and allow
683 the whole thing to run via HTTP.
685 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
687 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
690 * smoker on k81 fetching from k75 to verify cascading works. See
691 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
693 * maybe the loop should wait for CHECKSUMS file after every upload. And
694 CPAN.pm needs to deal with timestamps in the future.
696 * do not forget the dirtymark!
698 Text: have a new flag on recentfiles with the meaning: if this
699 changes, you're required to run a full rsync over all the files. The
700 reason why we set it would probably be: some foul happened. we injected
701 files in arbitrary places or didn't inject them although they changed.
702 The content of the flag? Timestamp? The relation between the
703 recentfiles would have to be inheritance from the principal, because any
704 out of band changes would soon later propagate to the next recentfile.
706 By upping the flag often one can easily ruin the slaves.
708 last out of band change? dirtymark?
710 Anyway, this implies that we read a potentially existing recentfile
713 And it implies that we have an eventloop that keeps us busy in 2-3
714 cycles, one for current stuff (tight loop) and one for the recentfiles
715 (cascade when principal has changed), one for the old stuff after a
718 And it implies that the out-of-band change in any of the recentfiles
719 must have a lock on the principal file and there is the place to set the
722 * start a FAQ, especially quick start guide questions. Also to aid those
723 problematic areas where we have no good solution, like the "links"
726 * wish feedback when we are slow.
730 * Remove a few DEBUG statements.
732 * The multiple-rrr way of doing things needs a new option to rmirror,
733 like piecemeal or so. Not urgent because after the first pass through,
734 things run smoothely. It's only ugly during the first pass.
736 * I have the suspicion that the code is broken that decides if the
737 neighboring RF needs to be seeded. I fear when too much time has gone
738 between two calls (in our case more than one hour), it would not seed
739 the neighbor. Of course this will never be noticed, so we need a good
742 * local/localroot confusion: I currently pass both options but one must
745 * accounts for early birds on PAUSE rsync daemon.
747 * hardcoded 20 seconds
749 * who mirrors the index? DOING now.
751 * which CPAN mirrors offer rsync?
753 * visit all XXX, visit all _float places
755 * rename the pathdb stuff, it's too confusing. No idea how.
757 * rrr-inotify, backpan, rrr-register
759 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
761 * current bugs: the pathdb seems to get no reset, the seeding of the
762 secondaryttl stuff seems not to have an effect. Have helped myself with
763 a rand(10), need to fix this back. So not checked in. Does the rand
766 The rand thing helps. The secondaryttl stuff was in the wrong line,
769 The pathdb stuff was because I called either _pathdb or __pathdb on the
770 wrong object. FIXED now.
772 * It's not so beautiful if we never fetch the recentfiles that are not
773 the principal, even if this is correct behaviour. We really do not need
774 them after we have fetched the whole content.
776 OK, we want a switch for that: secondaryttl DONE
778 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
780 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
781 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
782 it lets 40 follow. FIXED
784 * See that the long running process really only updates the principal
785 file unless it has missed a timespan during which something happened. If
786 nothing happened, it must notice even when it misses the timespan. DONE
788 * we must throw away the pathdb when we have reached the end of Z. From
789 that moment we can have a very small pathdb because the only reason for
790 a pathdb is that we know to ignore old records in old files. We won't
791 need this pathdb again before the next full pass over the data is
792 necessary and then we will rebuild it as we go along. DONE
794 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
796 * I think, Done::register_one is doing wrong in that it does not
797 conflate neighboring pieces. The covered() method cannot do this because
798 it has no recent_events array at hand. But register_one has it and could
799 do it and for some reason misses to do it (sometimes).
801 This means that the three tests I just wrote can probably not survive
802 because they test with an already broken Done structure.
804 The art now is to detect how it happens, then to reproduce, then write a
807 So from the logfile this is what happens: we have a good interval with
808 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
809 top of it. Locally we now mirror F2 and open a new done interval for it.
810 Then we mirror F1 but this time with the timestamp T1b. And when we then
811 try to close the gap, we do not find T1 but instead something older. We
812 should gladly accept this older piece and this would fix this bug.
816 * bug to fix: when the 1h file changes while rmirror is running, we do
817 correctly sync the new files but never switch to the 6h file but rather
818 stay in a rather quick loop that fetches the 1h file again and again.
820 Is it possible that we initialize a new object? Or does
821 get_remote_recentfile_as_tempfile overwrite something in myself?
823 Want a new option: _runstatusfile => $file which frequently dumps the
824 state of all recentfiles to a file.
828 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
830 * Todo: now teach update to verify the timestamp is about to write
831 against the previous and use _increase_a_bit if it doesn't comply with
832 strict monotony. DONE
834 * The problem of rounding. So far perl's default precision was
835 sufficient. One day it won't be. FakeFloat has an easy job when it is
836 only reading and other machines have written correctly. But when we want
837 to write a floating point number that is a bit larger than the other
838 one, then we need our own idea of precision.
840 Slaven said: just append a "1". This might be going towards the end of
841 usability too quickly. I'd like something that actually uses the decimal
842 system. Well, appending a 1 also does this but...
844 E.g. we have 1.0. nextup on this architecture is starting with
845 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
846 taken the 1.0000000000000003 and the next user comes and the time tells
847 him 1.0 again. He has to beat my number without stepping over the
848 nextup. This is much less space than I had when I chose 1,2,3.
850 What is also irritating is that nextup is architecture dependent. The
851 128 bit guy must choose very long numbers to fit in between whereas the
852 other one with 16 bit uses larger steps. But then the algorithm is the
853 same for both, so that would be a nice thing.
855 I see two situation where we need this. One is when Time::HiRes returns
856 us a value that is <= the last entry in our recentfile. In this case
857 (let's call it the end-case) we must fill the region between that number
858 and the next higher native floating point number. The other is when we
859 inject an old file into an old recentfile (we would then also set a new
860 dirtymark). We find the integer value already taken and need a slightly
861 different one (let's call it the middle-case). The difference between
862 the two situations is that the next user will want to find something
863 higher than my number in the end-case and something lower than my number
866 So I suggest we give the function both a value and an upper bound and it
867 calculates us a primitive middle. The upper bound in the middle-case is
868 the next integer. The upper bound on the end-case is the nextup floating
869 point number. But the latter poses another problem: if we have occupied
870 the middle m between x and nextup(x), then the nextup(m) will probably
871 not be the same as nextup(x) because some rounding will take place
872 before the nextup is calculated and when the rounding reaches the
873 nextup(x), we will end up at nextup(nextup(x)).
875 So we really need to consider the nextup and the nextdown from there and
876 then the middle and that's the number we may approach asymptotically.
879 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
881 * consider deprecating the use of RECENT.recent as a symlink. It turns
882 out to need extra hoops with the rsync options and just isn't worth it.
883 Or maybe these extra hoops are needed anyway for the rest of the tree?
884 Nope, can't be the case because not all filesystems support symlinks.
886 But before doing the large step, I'll deprecate the call of
887 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
888 only introduced to resolve RECENT.recent and complicates the routine far
889 beyond what it deserves.
891 DONE. Won't deprecate RECENT.recent, just moved its handling to the
894 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
896 * I think it's a bug that the rsync_option links must be set to true in
897 order to support RECENT.recent and that nobody cares to set it
898 automatically. Similar for ignore_link_stat_errors.
900 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
902 * Todo: find all todos together and make a plan what is missing for a
905 - verifytree or something like that. fsck maybe.
907 - rersyncrecent, the script itself? What it do?
909 - a way to only mirror the recentfiles without mirroring the whole
910 remote system such that people can decide to mirror only partially see
911 also 2008-08-30. .shadow-xxx directory? this also needed for a
912 filesystem that is still incomplete and might need the mirrorfiles for
915 - long living objects that mirror again and again. Inject something
916 into ta, see how it goes over to tb.
918 - how do we continue filling up the DONE system when we use an object
919 for the second time? "fully covered" and "uptodate" or new terminology.
921 - overview called on the wrong file should be understandable
923 - the meta data field that must change when we fake something up so that
924 the downstream people know they have to re-fetch everything.
926 - how tolerant are we against missing files upstream? how do we keep
927 track? there are legitimate cases where we did read upstream index right
928 before a file got deleted there and then find that file as new and want
929 it. There are other cases that are not self healing and must be tracked
932 - how, exactly, do we have to deal with deletes? With rsync errors?
934 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
935 authors) failed: No such file or directory (2)
937 The file above is a delete in 1h and a new in file 1M and the
938 delete in the locally running rmirror did not get propagated to the 1M
939 object. Bug. And the consequence is a standstill.
941 It seems that a slave that works with a file below the principal needs
942 to merge things all the way up to get rid of later deletes. Or keep
943 track of all deletes and skip them later. So we need a trackdeletes.pm
944 similar to the done.pm?
946 see also 2008-08-20 about spurious deletes that really have no add
947 counterpart and yet they are not wrong.
949 - consider the effect when resyncing the recentfile takes longer than
950 the time per loop. Then we never rsync any file. We need to diagnose
951 that and force an increase of that loop time. But when we later are fast
952 enough again because the net has recovered, then we need to switch back
953 to original parameters. ERm, no, it's enough to keep syncing at least
954 one file before refetching an index file.
956 - remember to verify that no temp files are left lying around and the
959 - status file for not long running jobs that want to track upstream with
962 - revisit all XXX _float areas and study Sub::Exporter DONE
964 - persistent DB even though we just said we do not need it. Just for
965 extended capabilities and time savings when, for example, upstream
966 announces a reset and we get new recentfiles and could then limit
967 ourselves to a subset of files (those that have a changed epoch) in a
968 first pass and would only then do the loop to verify the rest. Or
971 * Todo: aggregate files should know their feed and finding the principal
972 should be done stepwise. (?)
974 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
975 left locks around. DONE
977 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
979 * maybe extend the _overview so that it always says if and where the
980 last file is in the next file and where the next event in the next rf
981 would lie. No, don't like this anymore. REJECT
983 * take the two new redundant tests out again, only the third must
986 * Todo: add a sanity check if the merged structure is really pointing to
987 a different rf and that this different rf is larger. DONE
989 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
991 * now test, if they are overlapping. And test if there is a file in the
992 next rf that would fit into this rf's interval.
994 1h 1222324012.8474 1222322541.7963 0.4086
995 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
996 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
997 1W 1222313218.3626 1221708477.5829 167.9835
999 I suspect that somebody writes a merged timestamp without having merged
1000 and then somebody else relies on it.
1002 If aggregate is running, the intervals must not be extravagated, if it
1003 is not running, there must not be bounds, the total number of events in
1004 the system must be counted and must be controlled throughout the tests.
1005 That the test required the additional update was probably nonsense,
1006 because aggregate can cut pieces too. FIXED & DONE
1008 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1010 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1013 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1015 * Most apparent bug at the moment is that the recentfiles are fetched
1016 too often. Only the principal should be fetched and if it has not
1017 changed, the others should not be refetched. ATM I must admit that I'm
1018 happy that we refetch more often than needed because I can more easily
1019 fix bugs while the thing is running.
1021 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1022 done but the done system does not know about it. The reason for the
1023 failure is not known and we never reach the status uptodate because of
1024 this. We must get over it.
1026 Later it turns out that the origin server had a bug somewhere.
1027 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1028 array of the recentfile one position above. The bug was my own.
1030 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1032 * There is the race condition where the server does a delete and the
1033 slave does not yet know and then tries to download it because he sees
1034 the new. So for this time window we must be more tolerant against
1035 failure. If we cannot download a file, we should just skip it and should
1036 not retry immediately. The whole system should discover the lost thing
1037 later. Keeping track with the DONE system should really be a no brainer.
1039 But there is something more: the whole filesystem is a database and the
1040 recentfiles are one possible representation of it. It's a pretty useful
1041 representation I think that's why I have implemented something around
1042 it. But for strictly local operation it has little value. For local
1043 operation we would much rather have a database. So we would enter every
1044 recentfile reading and every rsync operation and for every file the last
1045 state change and what it leads to. Then we would always ignore older
1046 records without the efforts involved with recentfiles.
1048 The database would have: path,recentepoch,rsyncedon,deletedon
1050 Oh well, not yet clear where this leads to.
1052 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1054 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1055 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1057 * statusfile, maybe only on demand, alone to have a sharp debugging
1058 tool. It is locked and all recentfiles dump themselves into it and we
1059 can build a viewer that lets us know where we stand and what's inside.
1061 * remember: only the principal recentfile needs expiration, all others
1062 shall be expired by principal if it discovers that something has move
1065 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1067 * Always check if we stringify to a higher value than in the entry
1070 * And in covered make an additional check if we would be able to see a
1071 numerical difference between the two numbers and if we can't then switch
1072 to a different, more expensive algorithm. Do not want to be caught by
1073 floating surprises. DONE
1075 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1077 * caching has several aspects here: we can cache the interval of the
1078 recentfile which only will change when the mtime of the file changes. We
1079 must re-mirror the recentfile when its ttl has expired. Does have_read
1080 tell you anything? It counts nothing at all. Only the mtime is
1081 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1082 thing the have_mirrored because it is unrelated to the mtime.
1084 * Robustness of floating point calculations! I always thought that the
1085 string calculated by the origin server for the floating representation
1086 of the epoch time is just a string. When we convert it to a number and
1087 later back to a string, the other computer might come to a different
1088 conclusion. This must not happen, we want to preserve it under any
1089 circumstances. I will have to write tests with overlong sequences that
1090 get lost in arithmetic and must see if all still works well. DONE
1092 But one fragile point remains: if one host considers a>b and the other
1093 one considers them == but no eq. To prevent this, we must probably do
1094 some extra homework. DONE
1096 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1098 * the concept of tracking DONE needs an object per recentfile that has
1099 something like these methods:
1101 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1103 covered() register() covered()
1105 The unclear thing is how we translate points in time into intervals. We
1106 could pass a reference to the current recent_events array when running
1107 we_have(xxx) and let the DONE object iterate over it such that it only
1108 has to store a list of intervals that can melt into each other. Ah, even
1109 passing the list together with a list of indexes seems feasiable.
1111 Or maybe ask for the inverted list?
1113 Whenever the complete array is covered by the interval we say we are
1114 fully covered and if the recentfile is not expired, we are uptodate.
1116 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1118 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1120 * need a way to "return" the next entry after the end of a list. When
1121 the caller says "before" or "after" we would like to know if he could
1122 cover that interval/threshold or not because this influences the effect
1123 of a newer timestamp of that recentfile. DONE with $opt{info}.
1125 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1127 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1129 loop implies tixu (?). Nope, something like --statefile decides. Per
1132 T test, I init, X including removals, U nodirtymark
1134 So we have no concept of dirtymarks, we only trust that since we are
1135 running we have observed everything steadily. But people will not let
1136 this program run forever so we must consider both startup penalty and
1137 book keeping for later runs. We keep this for later. For now we write a
1138 long running mirror that merges several intervals.
1140 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1142 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1145 * rersyncrecent, the script: default to one week. The name of the switch
1146 is --after. Other switches? --loop!
1148 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1150 * need a switch --skip-deletes (?)
1152 * need a switch --enduser that tells us that the whole tempfile
1153 discipline is not needed when there is no downstream user. (?)
1155 Without this switch we cannot have a reasonable recent.pl that just
1156 displays the recent additions. Either we accept to download everything.
1157 Or we download temporary files without the typical rsync protocol
1160 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1161 File::Temp::tempdir, this might be a win.
1163 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1165 * apropos missing: we have no push, we never know the downstream
1166 servers. People who know their downstream hosts and want to ascertain
1167 something will want additional methods we have never thought about, like
1168 update or delete a certain file.
1170 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1172 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1173 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1176 * rsync options: it's a bit of a pain that we usually need several rsync
1177 options, like compress, links, times, checksum and that there is no
1178 reasonable default except the original rsync default. I think wee can
1179 safely assume that the rsync options are shared between all recentfile
1180 instances within one recent tree.
1182 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1184 * deletes: if a delete follows an add quickly enough it may happen that
1185 a downstream mirror did not see the add at all! It seems this needs to
1186 be mentioned somewhere. The point here is that even if the downstream is
1187 never missing the principal timeframe it may encounter a "delete" that
1188 has no complimentary "add" anywhere.
1190 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1192 * I suspect the treat of metadata is incorrect during read or something.
1193 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1194 contained more than 6 hours worth of data. At 06:08 we merged into the
1195 1d file. We need to take snapshots of the 6h file over the course of an
1196 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1199 Much worse: watching the 1h file: right at the moment (at 06:35) it
1200 covers 1218867584-1219120397 which is 70 hours.
1202 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1203 08:19:44 2008, that is when I checked out last time, so it seems to be
1204 aggregating and never truncating?
1206 No, correct is: it is never truncating; but wrong is: it is aggregating.
1207 It does receive a lot of events from time to time from a larger file.
1208 Somehow a large file gets merged into the small one and because the
1209 "meta/merged" attribute is missing, nobody is paying attention. I
1210 believe that I can fix this by making sure that metadata are honoured
1211 during read. DONE and test adjusted.
1213 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1215 * grand renaming plan
1217 remotebase => remoteroot to fit well with localroot DONE
1218 local_path() => localroot seems to me should already work DONE
1219 recentfile_basename => rfilename no need to stress it has no slash DONE
1221 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1222 kern? It would anyway need a deprecation cycle because it is an
1223 important constructor.
1225 * I like the portability that Data::Serializer brings us but the price
1226 is that some day we might find out that it is slowing us a bit. We'll
1229 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1231 * should we not enter the interval of the principal (or the interval of
1232 the merging file?) in every aggregated/merged file?
1234 * we should aim at a first release and give up on thinking about
1235 sanitizing stuff and zloop. Let's just admit that a full traditional
1236 rsync is the only available sanitizer ATM. Otherwise it's complicated
1237 stuff: sanitizing on the origin server, sanitizing on the slaves,
1238 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1239 get the basics out before this becomes a major cause for mess.
1241 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1243 * On OSes not supporting symlinks we expect that RECENT.recent contains
1244 the contents of the principal recentfile. Actually this is identical on
1245 systems supporting symlinks. Simple, what follows from that is that we
1246 need to keep the serializer in the metadata because we cannot read it
1247 from the filename, doesn't it? Of course not. It's a chicken and egg
1248 problem. This leaves us with the problem to actually parse the
1249 serialized data to find out in which format it is. So who can do the 4
1250 or 5 magics we wanted to support? File::LibMagic?
1252 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1254 * remotebase and recentfile_basename are ugly names. Now that we need a
1255 word for the shortest/principal/driving recentfile too we should do
1258 localroot is good. rfile is good. local_path() is bad, local_path($path)
1259 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1262 Up to now remotebase was the string that described the remote root
1263 directory in rsync notation, like pause.perl.org::authors. And
1264 recentfile_basename was "RECENT-1h.yaml".
1266 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1268 * The test that was added in today's checkin is a good start for a test
1269 of rmirror. We should have more methods in Recent.pm: verify,
1270 addmissingfiles. We should verify the current tree, then rmirror it and
1271 then verifytree the copy. We could then add some arbitrary file and let
1272 it be discovered by addmissingfiles, then rmirror again and then
1273 verifytree the copy again.
1275 Then we could start stealing from csync2 sqlite database [no port to
1276 OSX!] and fill a local DB. And methods to compare the database with the
1277 recentfiles. Our strength is that in principle we could maintain state
1278 with a single float. We have synced up to 1234567890.123456. If the Z
1279 file does not add new files all we have to do is mirror the new ones and
1282 This makes it clear that we should extend current protocol and declare
1283 that we cheat when we add files too late, just to help the other end
1284 keeping track. Ah yes, that's what was meant when zloop was mentioned
1287 Maybe need to revisit File::Mirror to help me with this task.
1289 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1291 * There must be an allow-me-to-truncate flag in every recentfile.
1292 Without it one could construct a sequence of updates winning the locking
1293 battle against the aggregator. Only if an aggregator has managed to
1294 merge data over to the next level, truncating can be allowed. DONE with
1297 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1299 * We should probably guarantee that no duplicates enter the aggregator
1302 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1304 * To get merge operation faster would need a good benchmark test. What
1305 02 spits out isn't reliable enough and is dominated by many other
1308 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1309 Author: Andreas J. Koenig <k@k75.(none)>
1310 Date: Sat Aug 2 07:58:04 2008 +0200
1314 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1315 Author: Andreas J. Koenig <k@k75.(none)>
1316 Date: Sat Aug 2 11:40:29 2008 +0200
1318 there was a lot of trying but the effect is hardly measurable with
1321 * overhead of connecting seems high. When setting
1322 max_files_per_connection to 1 we see that.
1324 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1326 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1328 25138.2215600014/3600 = 6.98283932222261
1330 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1331 be a bug in the aggregator. FIXED
1333 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1335 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1336 we have it, can we mirror it, mirror it!
1338 I fear this needs a new class which might be called
1339 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1340 and treat them as an entity. I realize that a single recentfile may be
1341 sufficient for certain tasks and that it is handy for the low level
1342 programmer but it is not nice to use. If there is a delete in the 1h
1343 file then the 6h file still contains it. Seekers of the best information
1344 need to combine at least some of the recentfiles most of the time.
1346 There is the place for the Z loop!
1348 But the combination is something to collect in a database, isn't it. Did
1349 csync2 just harrumph?
1351 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1353 * it just occurred to me that hosts in the same mirroring pool could
1354 help out each other even without rewriting the recentfile. Just fetch
1355 the stuff to mirror from several places, bingo. But that's something
1356 that should rather live in a separate package or in rsync directly.
1358 * cronjobs are unsuited because with ntp they would all come at the full
1359 minute and disturb each other. Besides that I'd hate to have a backbone
1360 with more than a few seconds latency.
1362 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1364 * a second rsync server with access control for PAUSE. Port? 873 is the
1365 standard port, let's take 8873.
1367 * if there were a filesystem based on this, it would have a slow access
1368 to inexistent files. It would probably provide wrong readdir (only based
1369 on current content) or also a slow one (based on a recentfile written
1370 after the call). But it would provide fast access to existing files. Or
1371 one would deliberately allow slightly blurred answers based on some
1372 sqlite reflection of the recentfiles.
1374 * todo: write a variant of mirror() that combines two or more
1375 recentfiles and treats them like one
1377 * todo: signal handler to remove the tempfile
1379 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1381 * now that we have the symlink I forgot how it should be used in
1384 * the z loop: add missing files to Z file. Just append them (instead of
1385 prepending). So one guy prepends something from the Y file from time to
1386 time and another guy appends something rather frequently. Collecting
1387 pond. When Y merges into Z, things get epoch and the collecting pond
1388 gets smaller. What exactly are "missing files"?
1390 take note of current epoch of the alpha file, let's call it the
1393 find all files on disk
1395 remove all files registered in the recentworld up to recent-ts
1397 remove all files that have been deleted after recent-ts according to
1400 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1402 * rersyncrecent might be a cronjob with a (locked) state file which
1403 contains things like after and maybe last z sync or such?
1405 rrr-mirror might be an alternative name but how would we justify the
1406 three Rs when there is no Re-Rsync-Recent?
1408 With the --loop parameter it is an endless loop, without it is no loop.
1409 At least this is simple.
1411 * todo: new accssor z-interval specifies how often the Z file is updated
1412 against the filesystem. We probably want no epoch stamp on these
1413 entries. And we want to be able to filter the entries (e.g. no
1414 by-modules and by-category tree)
1416 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1418 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1419 into Z. We do not need the other files filled up with filesystem
1422 * need interface to query for a file in order to NOT call update on
1423 PAUSE a second time within a short time.
1425 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1427 * recommended update interval? Makes no sense, is different for
1434 change-log-default-name: "Todo"