1 2009-06-22 Andreas J. Koenig <andk@cpan.org>
3 * Todo: rrr-init to generate some index files so that rrr-server can be
6 2009-05-02 Andreas J. Koenig <andk@cpan.org>
8 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
9 so we can use it for normal update with a single file. And add a cronjob
10 that runs rrr-fsck so we get a reminder when something is not working
13 2009-04-27 Andreas J. Koenig <andk@cpan.org>
15 * Todo: I still see more link_stat errors than expected. I come back to
16 the former suspicion that we do not do necessary bookkeeping on deleted
17 files so that when we reach the "new" event on a meanwhile deleted file
18 we try to mirror it although we could know better. It's just noise, no
19 harm but noise is irritating.
21 * Todo: more experimenting with runstatusfile to get it official
23 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
26 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
28 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
30 And that ignore_link_stat_error is propagated in sparse_clone.
34 For the record: the bug was that ignore_link_stat_error was lost during
35 sparse_clone. And then the second bug was that it did not default to
36 true. Both are now fixed.
38 2009-04-25 Andreas J. Koenig <andk@cpan.org>
40 * Todo: fill rrr program with life
42 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
44 * Minor bug: I see complaints about files not existing like
45 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
46 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
48 In the RECENT files these are delete events but of course also new
49 events in older recent files.
51 I think we do not properly remember deletes when we skip-deletes and so
52 do not filter them out when we later see the "new" event.
54 Yes, but that what ignore_link_stat_error is for, we know about this
55 race condition. We must make ignore_link_stat_error default to true.
59 * bug in Done.pm failing to merge two fields into one when a third is
60 present. FIXED and accompanied with new test.
62 2009-04-24 Andreas J. Koenig <andk@cpan.org>
64 * Todo: keep the index clean and run some kind of nondestructive fsck 4
67 * who is our backbone?
69 http://cpan.cpantesters.org/authors/02STAMP # barbie
70 http://cpan.solfo.com/authors/02STAMP # abh
71 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
73 2009-04-17 Andreas J. Koenig <andk@cpan.org>
75 * the changelog helper I included in the Makefile release memo does not
76 show tags and is much less useful than I thought. gitk is probably much
79 2009-04-16 Andreas J. Koenig <andk@cpan.org>
83 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
84 #### Ydr_.yaml ... DONE
85 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
86 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
87 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
88 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
89 #### gone already at cpan-pause.pl line 0
91 Running without skip-deletes now but cannot reproduce.
93 2009-04-15 Andreas J. Koenig <andk@cpan.org>
95 * consider state directory such we can restart after a ^C or come again
98 Reuse _runstatusfile: write atomically, see how we can use the dump to
99 start running again. Do not leak and remove globs before writing. Do not
100 only dump one rf, dump the whole complex.
102 * consider the effect of ^C. Is it safe?
104 * want to use Perl::Version to up the versions on every release. Only if
105 things have changed, of course. Does this actually work?
107 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
108 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
110 Must do the bookkeeping myself when this should happen, maybe based on git?
112 I think I prefer the same version number for all pm files. But one also
113 needs a routine for setting them. Ahh:
115 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
121 cpan[2]> m /Mirror::Recent/
122 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
123 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
124 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
125 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
128 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
130 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
132 I just tried MM 6.5102 and it again produces only a provides for Recent,
133 not for the other modules.
135 My own fault. FIXED in the Makefile.PL.
137 2009-04-13 Andreas J. Koenig <andk@cpan.org>
139 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
141 * is it true (as stated at
142 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
143 that rsync can lead to errors when upstream changes before client sync
144 has completed? Can we see the error when we run this on upstream:
147 use Time::HiRes qw(time);
149 open my $fh, ">", "changingfile.txt.new" or die;
150 print $fh (time."\n") x 1000000;
152 rename "changingfile.txt.new", "changingfile.txt" or die;
156 And on the receiving end:
159 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
163 I cannot get it to fail. Apparently it is sufficient when upstream
164 always writes atomically (which of course is mandatory)?
166 SENT to mailing list Apr 28.
170 > https://fedorahosted.org/InstantMirror/
171 > https://www.redhat.com/mailman/listinfo/instantmirror-list
172 > irc.freenode.net channel #instantmirror
174 A cool name for a project. Inspires me to write a few sentences about
175 bittorrent's role in the grand picture.
177 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
178 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
180 2009-04-12 Andreas J. Koenig <andk@cpan.org>
182 * Interesting last minute bug during real download testing: the output
183 isn't as pretty anymore, it seems that more work is being done than
186 Introducing a highlevel _alluptodate method helped a lot in debugging
189 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
190 different on 1h file than on the other files. This lead to frequent
191 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
194 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
196 * Todo: try if we can be faster with a native float package. I'm really
197 glad to have a machine exposing perl floating point bugs, it forced me
198 to retract this item from the changes file: "several speedups in the
199 fakebigfloat code" before I get cpan testers fails.
201 2009-04-11 Andreas J. Koenig <andk@cpan.org>
203 * Bug: when the dirtymark changes upstream, then the downstream server
204 notices it immediately and mirrors the "1h" file but then leaves the
205 rmirror loop and continues with "6h" on the next round through. This
206 behaviour goes on until it reaches the Z file. Because we eat the Z file
207 piecemeal we leave the loop after a while but after that everybody seems
208 to have forgotten that there is some work left to be done. Bad, bad,
211 ETOOCONFUSING better name?
213 Recentfile::get_remote_recentfile_as_tempfile OK
214 Recentfile::resolve_recentfilename split_rfilename DONE
216 Recent::_principal_recentfile_object _principal_recentfile_fromremote
217 Recent::principal_recentfile OK
218 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
220 some void, some not, some not void but called in void context.
222 2009-04-10 Andreas J. Koenig <andk@cpan.org>
224 * Bug? When the dirtymark changes then the second tier hosts quickly
225 reset their DONE state and restart mirroring. But then they mirror
226 potentially outdated index files with inconsistent dirtymark because the
227 first tier box may be a bit behind on the large recentfiles.
229 This means we need a brake when iterating over the recentfiles that
230 refuses to use a recentfile with the wrong dirtymark.
232 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
234 * install the bin/* files and talk about rrr-overview somewhere.
236 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
237 created outside the target tree.
239 * Barbie asks for better logging capabilities
241 2009-03-29 Andreas J. Koenig <andk@cpan.org>
243 * known bug: cannot configure the Recentfile objects to keep delete
244 events. How to call the accessor for that? keep_delete_objects_forever?
246 Sounds acceptable. FIXED (but untested).
248 * known bug: keeps temporary index files lying around.
250 2009-03-24 Andreas J. Koenig <andk@cpan.org>
252 * Some equivalent for
254 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
258 2009-03-22 Andreas J. Koenig <andk@cpan.org>
260 * k81 is again client of k75 with different parameters.
262 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
263 about it all the time and nobody reacts accordingly; (2) lots of temp
264 files get created and not removed; culprit the new call to
265 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
266 caller has to remove the tempfile after use.
268 Need the drawing board.
270 PARTLY FIXED: retracting the idea to call
271 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
272 and unseed such that uptodate now means "we have mirrored this
273 recentfile's bunch of files" and seeded means "somebody believes the
274 index file for this rf needs be refreshed". This seems to work now.
276 Still collecting temp files that nobody cares to depollute.
278 2009-03-21 Andreas J. Koenig <andk@cpan.org>
280 * Bug?: should it be harder than it is atm to set the timestamp to the
283 * bug with native integers:
287 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
291 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
294 Native integer broke when native math was turned off. FIXED
296 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
297 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
300 Yes, records out of order:
303 447004 epoch: 995885533
304 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
307 447008 epoch: 995890358
308 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
311 447012 epoch: 995892221
312 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
315 FIXED with sanity check and later with the integer fix.
317 * Bug: want the index files in a .recent directory
319 * Bug: lots of dot files are not deleted in time
321 * possible test case: can a delete change the timestamp? This would
322 probably break the order of events.
324 2009-03-20 Andreas J. Koenig <andk@cpan.org>
326 * 1233701831.34486 what's so special about this number/string? It
328 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
329 atm lives in Y,Q, and Z.
331 It is the first entry after id/--skip-locking which has timestamp
332 1234164228.11325 which represents a file that doesn't exist anymore.
336 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
337 _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
338 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
339 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
340 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
341 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
342 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
343 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
344 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
345 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
346 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
347 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
348 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
350 FIXED, it was the "--skip-locking" file where manual intervention was participating
352 * bug on the mirroring slave: when the dirtymark gets increased we
353 probably do not reset the done intervals. The mirrorer stays within
354 tight bounds where it tries to sync with upstream and never seems to
355 finish. In the debugging state file I see lots of identical intervals
356 that do not get collapsed. When I restart the mirrorer it dies with:
358 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
359 _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
360 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
361 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
362 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
363 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
364 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
365 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
366 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
367 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
369 and debugging stands at
375 epoch: 1237400802.5789
376 path: id/J/JO/JOHND/CHECKSUMS
379 epoch: 1237400807.97514
380 path: id/M/MI/MIYAGAWA/CHECKSUMS
383 epoch: 1237400817.94363
384 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
392 and it is reproducable.
394 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
395 pause has a fresh one.
397 FIXED, it was the third anded term in each of the ifs in the IV block in
398 _register_one: with that we make sure that we do not stamp on valuable
401 2009-03-17 Andreas J. Koenig <andk@cpan.org>
403 * done: verified the existence of the floating point bug in bleadperl
404 and verified that switching from YAML::Syck to YAML::XS does not resolve
407 BTW, the switch was doable with
409 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
411 and should be considered as a separate TODO
413 * todo: integrate a dirty update with two aggregate calls before
414 unlocking for frictionless dirtying DONE
416 * todo: start the second rsync daemon on pause DONE
418 * todo: move index files to .recent: this cannot simply be done by
419 setting filenameroot to .recent/RECENT. Other parts of the modules rely
420 on the fact that dirname(recentfile) is the root of the mirrored tree.
422 2009-03-16 Andreas J. Koenig <andk@cpan.org>
424 * What was the resolution of the mirror.pl delete hook bug? Do we call
425 the delete hook when pause removes a file from MUIR?
427 * Today on pause: Updating 2a13fba..29f284d and installing it for
428 /usr/local/perl-5.10.0{,-RC2}
430 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
431 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
433 2009-03-15 Andreas J. Koenig <andk@cpan.org>
435 * currently recent_events has the side effect of setting dirtymark
436 because it forces a read on the file. That should be transparent, so
437 that the dirtymark call always forces a cache-able(?) read.
439 * The bug below is -- after a lot of trying -- not reproducible on a
440 small script, only in the large test script. The closest to the output
446 my $x = "01237123229.8814";
451 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
453 $l = "1237123231.22458";
454 $r = "1237123231.22458";
456 Devel::Peek::Dump $l;
457 Devel::Peek::Dump $r;
458 Devel::Peek::Dump $x = $l <=> $r;
462 The checked in state at c404a85 fails the test with my
463 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
464 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
465 in the test script always consistent.
467 * Todo: write a test that inserts a second dirty file with an already
468 existing timestamp. DONE
470 * Bug in perl 5.10 on my 64bit box:
472 DB<98> Devel::Peek::Dump $l
473 SV = PVMG(0x19e0450) at 0x142a550
475 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
477 NV = 1237123231.22458
478 PV = 0x194ce70 "1237123231.22458"\0
482 DB<99> Devel::Peek::Dump $r
483 SV = PVMG(0x19e0240) at 0x142a3e8
485 FLAGS = (PADMY,POK,pPOK)
488 PV = 0x19ff900 "1237123231.22458"\0
492 DB<100> Devel::Peek::Dump $l <=> $r
493 SV = IV(0x19ea6e8) at 0x19ea6f0
495 FLAGS = (PADTMP,IOK,pIOK)
498 DB<101> Devel::Peek::Dump $l
499 SV = PVMG(0x19e0450) at 0x142a550
501 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
503 NV = 1237123231.22458
504 PV = 0x194ce70 "1237123231.22458"\0
508 DB<102> Devel::Peek::Dump $r
509 SV = PVMG(0x19e0240) at 0x142a3e8
511 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
513 NV = 1237123231.22458
514 PV = 0x19ff900 "1237123231.22458"\0
518 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
520 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
522 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
524 * $obj->merge ($other) needs to learn about equal epoch which may happen
525 since dirty_epoch intruded.
527 * Wontfix anytime soon: I think we currently do not support mkdir. Only
530 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
532 * Todo: continue working on update(...,$dirty_epoch). It must be
533 followed by a fast_aggregate! DONE
535 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
537 * maybe we need a closest_entry or fitting_interval or something like
538 that. We want to merge an event into the middle of some recentfile.
539 First we do not know which file, then we do not know where to lock,
540 where to enter the new item, when and where to correct the dirtymark.
542 So my thought is we should first find which file.
544 Another part of my brain answers: what would happen if we would enter
545 the new file into the smallest file just like an ordinary new event,
546 just as an old event?
548 (1) we would write a duplicate timestamp? No, this would be easy to
551 (2) we would make the file large quickly? Yes, but so what? We are
552 changing the dirtymark, so are willing to disturb the downstream hosts.
554 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
556 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
560 https://rt.cpan.org/Ticket/Display.html?id=41199
562 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
564 Strangest output being HAKANARDO who managed to upload
566 Here is a better oneliner that includes also the first line of each
569 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
571 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
573 * memory leak in the syncher? It currently weighs 100M.
577 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
580 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
582 * lookup by epoch and by path and use this ability on the pause to never
583 again register a file twice that doesn't need it. Let's call it
586 * after the dirtymark is done: fill up recentfiles with fake (historic)
587 entries; fill up with individual corrections; algorithm maybe to be done
588 with bigfloat so that we can always place something in the middle
589 between two entries. Before we must switch to bigfloat we could try to
590 use Data::Float::nextup to get the.
592 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
595 * dirtymark now lives in Recentfile, needs to be used in rmirror.
597 * find out why the downloader died after a couple of hours without a net
598 connection. Write a test that survives the not-existence of the other
601 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
603 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
604 it makes less sense with tight coupling for secondary files. But asking
605 the server what the current epoch is might be cheaper on HTTP than on
606 rsync. (Needs to be evaluated)
608 * remove the 0.00 from the verbose overview in the Merged column in the
611 * write tests that expose the problems of the last few days: cascading
612 client/server roles, tight coupling for secondary RFs, deletes after
615 * Some day we might want to have policy options for the slave:
616 tight/loose/no coupling with upstream for secondary RFs. tight is what
617 we have now. loose would wait until a gap occurs that can be closed.
619 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
621 * revisit all $rfs->[$i+1] places if they now make sense still
623 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
625 * another bug is the fact that the mirror command deletes files before
626 it unhides the index file, thus confusing downstream slaves. We must not
627 delete before unhiding and must delete after unhiding. FIXED.
629 * new complication about the slave that is playing a server role.
630 Currently we mirror from newest to oldest with a hidden temporary file
631 as index. And when one file is finished, we unhide the index file.
632 Imagine the cascading server/slave is dead for a day. It then starts
633 mirroring again with the freshest thing and unhides the freshest index
634 file when it has worked through it. In that moment it exposes a time
635 hole. Because it now works on the second recentfile which is still
638 We currently do nothing special to converge after such a drop out. At
639 least not intentionally and robustly and thought through.
641 The algorithm we use to seed the next file needs quite a lot of more
642 robustness than it currently has. Something to do with looking at the
643 merged element of the next rf and when it has dropped off, we seed
644 immediately. And if it ramains dropped off, we seed again, of course.
646 Nope, looking from smaller to larger RFS we look at the merged element
647 of this RF and at the minmax/max element of the next RF. If that
648 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
651 And we need a public accessor seed and unseed or seeded. But not the mix
652 of public and private stuff that then is used behind the back.
654 And then the secondary* stuff must go.
656 And we must understand what the impact is on the DONE system. Can it go
657 unnoticed that there was a hole? And could the DONE system have decided
658 the hole is covered? This should be testable with three directories where
659 the middle stops working for a while. Done->merge is suspicious, we must
660 stop it from merging non-conflatable neighbors due to broken continuity.
664 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
666 * Slaven suggests to have the current epoch or the whole current
667 recentfile available from the HTTP server and take it away with
668 keepalive. This direction goes the granularity down to subseconds.
670 We might want to rewrite everything to factor out transport and allow
671 the whole thing to run via HTTP.
673 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
675 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
678 * smoker on k81 fetching from k75 to verify cascading works. See
679 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
681 * maybe the loop should wait for CHECKSUMS file after every upload. And
682 CPAN.pm needs to deal with timestamps in the future.
684 * do not forget the dirtymark!
686 Text: have a new flag on recentfiles with the meaning: if this
687 changes, you're required to run a full rsync over all the files. The
688 reason why we set it would probably be: some foul happened. we injected
689 files in arbitrary places or didn't inject them although they changed.
690 The content of the flag? Timestamp? The relation between the
691 recentfiles would have to be inheritance from the principal, because any
692 out of band changes would soon later propagate to the next recentfile.
694 By upping the flag often one can easily ruin the slaves.
696 last out of band change? dirtymark?
698 Anyway, this implies that we read a potentially existing recentfile
701 And it implies that we have an eventloop that keeps us busy in 2-3
702 cycles, one for current stuff (tight loop) and one for the recentfiles
703 (cascade when principal has changed), one for the old stuff after a
706 And it implies that the out-of-band change in any of the recentfiles
707 must have a lock on the principal file and there is the place to set the
710 * start a FAQ, especially quick start guide questions. Also to aid those
711 problematic areas where we have no good solution, like the "links"
714 * wish feedback when we are slow.
718 * Remove a few DEBUG statements.
720 * The multiple-rrr way of doing things needs a new option to rmirror,
721 like piecemeal or so. Not urgent because after the first pass through,
722 things run smoothely. It's only ugly during the first pass.
724 * I have the suspicion that the code is broken that decides if the
725 neighboring RF needs to be seeded. I fear when too much time has gone
726 between two calls (in our case more than one hour), it would not seed
727 the neighbor. Of course this will never be noticed, so we need a good
730 * local/localroot confusion: I currently pass both options but one must
733 * accounts for early birds on PAUSE rsync daemon.
735 * hardcoded 20 seconds
737 * who mirrors the index? DOING now.
739 * which CPAN mirrors offer rsync?
741 * visit all XXX, visit all _float places
743 * rename the pathdb stuff, it's too confusing. No idea how.
745 * rrr-inotify, backpan, rrr-register
747 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
749 * current bugs: the pathdb seems to get no reset, the seeding of the
750 secondaryttl stuff seems not to have an effect. Have helped myself with
751 a rand(10), need to fix this back. So not checked in. Does the rand
754 The rand thing helps. The secondaryttl stuff was in the wrong line,
757 The pathdb stuff was because I called either _pathdb or __pathdb on the
758 wrong object. FIXED now.
760 * It's not so beautiful if we never fetch the recentfiles that are not
761 the principal, even if this is correct behaviour. We really do not need
762 them after we have fetched the whole content.
764 OK, we want a switch for that: secondaryttl DONE
766 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
768 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
769 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
770 it lets 40 follow. FIXED
772 * See that the long running process really only updates the principal
773 file unless it has missed a timespan during which something happened. If
774 nothing happened, it must notice even when it misses the timespan. DONE
776 * we must throw away the pathdb when we have reached the end of Z. From
777 that moment we can have a very small pathdb because the only reason for
778 a pathdb is that we know to ignore old records in old files. We won't
779 need this pathdb again before the next full pass over the data is
780 necessary and then we will rebuild it as we go along. DONE
782 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
784 * I think, Done::register_one is doing wrong in that it does not
785 conflate neighboring pieces. The covered() method cannot do this because
786 it has no recent_events array at hand. But register_one has it and could
787 do it and for some reason misses to do it (sometimes).
789 This means that the three tests I just wrote can probably not survive
790 because they test with an already broken Done structure.
792 The art now is to detect how it happens, then to reproduce, then write a
795 So from the logfile this is what happens: we have a good interval with
796 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
797 top of it. Locally we now mirror F2 and open a new done interval for it.
798 Then we mirror F1 but this time with the timestamp T1b. And when we then
799 try to close the gap, we do not find T1 but instead something older. We
800 should gladly accept this older piece and this would fix this bug.
804 * bug to fix: when the 1h file changes while rmirror is running, we do
805 correctly sync the new files but never switch to the 6h file but rather
806 stay in a rather quick loop that fetches the 1h file again and again.
808 Is it possible that we initialize a new object? Or does
809 get_remote_recentfile_as_tempfile overwrite something in myself?
811 Want a new option: _runstatusfile => $file which frequently dumps the
812 state of all recentfiles to a file.
816 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
818 * Todo: now teach update to verify the timestamp is about to write
819 against the previous and use _increase_a_bit if it doesn't comply with
820 strict monotony. DONE
822 * The problem of rounding. So far perl's default precision was
823 sufficient. One day it won't be. FakeFloat has an easy job when it is
824 only reading and other machines have written correctly. But when we want
825 to write a floating point number that is a bit larger than the other
826 one, then we need our own idea of precision.
828 Slaven said: just append a "1". This might be going towards the end of
829 usability too quickly. I'd like something that actually uses the decimal
830 system. Well, appending a 1 also does this but...
832 E.g. we have 1.0. nextup on this architecture is starting with
833 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
834 taken the 1.0000000000000003 and the next user comes and the time tells
835 him 1.0 again. He has to beat my number without stepping over the
836 nextup. This is much less space than I had when I chose 1,2,3.
838 What is also irritating is that nextup is architecture dependent. The
839 128 bit guy must choose very long numbers to fit in between whereas the
840 other one with 16 bit uses larger steps. But then the algorithm is the
841 same for both, so that would be a nice thing.
843 I see two situation where we need this. One is when Time::HiRes returns
844 us a value that is <= the last entry in our recentfile. In this case
845 (let's call it the end-case) we must fill the region between that number
846 and the next higher native floating point number. The other is when we
847 inject an old file into an old recentfile (we would then also set a new
848 dirtymark). We find the integer value already taken and need a slightly
849 different one (let's call it the middle-case). The difference between
850 the two situations is that the next user will want to find something
851 higher than my number in the end-case and something lower than my number
854 So I suggest we give the function both a value and an upper bound and it
855 calculates us a primitive middle. The upper bound in the middle-case is
856 the next integer. The upper bound on the end-case is the nextup floating
857 point number. But the latter poses another problem: if we have occupied
858 the middle m between x and nextup(x), then the nextup(m) will probably
859 not be the same as nextup(x) because some rounding will take place
860 before the nextup is calculated and when the rounding reaches the
861 nextup(x), we will end up at nextup(nextup(x)).
863 So we really need to consider the nextup and the nextdown from there and
864 then the middle and that's the number we may approach asymptotically.
867 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
869 * consider deprecating the use of RECENT.recent as a symlink. It turns
870 out to need extra hoops with the rsync options and just isn't worth it.
871 Or maybe these extra hoops are needed anyway for the rest of the tree?
872 Nope, can't be the case because not all filesystems support symlinks.
874 But before doing the large step, I'll deprecate the call of
875 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
876 only introduced to resolve RECENT.recent and complicates the routine far
877 beyond what it deserves.
879 DONE. Won't deprecate RECENT.recent, just moved its handling to the
882 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
884 * I think it's a bug that the rsync_option links must be set to true in
885 order to support RECENT.recent and that nobody cares to set it
886 automatically. Similar for ignore_link_stat_errors.
888 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
890 * Todo: find all todos together and make a plan what is missing for a
893 - verifytree or something like that. fsck maybe.
895 - rersyncrecent, the script itself? What it do?
897 - a way to only mirror the recentfiles without mirroring the whole
898 remote system such that people can decide to mirror only partially see
899 also 2008-08-30. .shadow-xxx directory? this also needed for a
900 filesystem that is still incomplete and might need the mirrorfiles for
903 - long living objects that mirror again and again. Inject something
904 into ta, see how it goes over to tb.
906 - how do we continue filling up the DONE system when we use an object
907 for the second time? "fully covered" and "uptodate" or new terminology.
909 - overview called on the wrong file should be understandable
911 - the meta data field that must change when we fake something up so that
912 the downstream people know they have to re-fetch everything.
914 - how tolerant are we against missing files upstream? how do we keep
915 track? there are legitimate cases where we did read upstream index right
916 before a file got deleted there and then find that file as new and want
917 it. There are other cases that are not self healing and must be tracked
920 - how, exactly, do we have to deal with deletes? With rsync errors?
922 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
923 authors) failed: No such file or directory (2)
925 The file above is a delete in 1h and a new in file 1M and the
926 delete in the locally running rmirror did not get propagated to the 1M
927 object. Bug. And the consequence is a standstill.
929 It seems that a slave that works with a file below the principal needs
930 to merge things all the way up to get rid of later deletes. Or keep
931 track of all deletes and skip them later. So we need a trackdeletes.pm
932 similar to the done.pm?
934 see also 2008-08-20 about spurious deletes that really have no add
935 counterpart and yet they are not wrong.
937 - consider the effect when resyncing the recentfile takes longer than
938 the time per loop. Then we never rsync any file. We need to diagnose
939 that and force an increase of that loop time. But when we later are fast
940 enough again because the net has recovered, then we need to switch back
941 to original parameters. ERm, no, it's enough to keep syncing at least
942 one file before refetching an index file.
944 - remember to verify that no temp files are left lying around and the
947 - status file for not long running jobs that want to track upstream with
950 - revisit all XXX _float areas and study Sub::Exporter DONE
952 - persistent DB even though we just said we do not need it. Just for
953 extended capabilities and time savings when, for example, upstream
954 announces a reset and we get new recentfiles and could then limit
955 ourselves to a subset of files (those that have a changed epoch) in a
956 first pass and would only then do the loop to verify the rest. Or
959 * Todo: aggregate files should know their feed and finding the principal
960 should be done stepwise. (?)
962 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
963 left locks around. DONE
965 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
967 * maybe extend the _overview so that it always says if and where the
968 last file is in the next file and where the next event in the next rf
969 would lie. No, don't like this anymore. REJECT
971 * take the two new redundant tests out again, only the third must
974 * Todo: add a sanity check if the merged structure is really pointing to
975 a different rf and that this different rf is larger. DONE
977 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
979 * now test, if they are overlapping. And test if there is a file in the
980 next rf that would fit into this rf's interval.
982 1h 1222324012.8474 1222322541.7963 0.4086
983 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
984 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
985 1W 1222313218.3626 1221708477.5829 167.9835
987 I suspect that somebody writes a merged timestamp without having merged
988 and then somebody else relies on it.
990 If aggregate is running, the intervals must not be extravagated, if it
991 is not running, there must not be bounds, the total number of events in
992 the system must be counted and must be controlled throughout the tests.
993 That the test required the additional update was probably nonsense,
994 because aggregate can cut pieces too. FIXED & DONE
996 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
998 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1001 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1003 * Most apparent bug at the moment is that the recentfiles are fetched
1004 too often. Only the principal should be fetched and if it has not
1005 changed, the others should not be refetched. ATM I must admit that I'm
1006 happy that we refetch more often than needed because I can more easily
1007 fix bugs while the thing is running.
1009 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1010 done but the done system does not know about it. The reason for the
1011 failure is not known and we never reach the status uptodate because of
1012 this. We must get over it.
1014 Later it turns out that the origin server had a bug somewhere.
1015 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1016 array of the recentfile one position above. The bug was my own.
1018 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1020 * There is the race condition where the server does a delete and the
1021 slave does not yet know and then tries to download it because he sees
1022 the new. So for this time window we must be more tolerant against
1023 failure. If we cannot download a file, we should just skip it and should
1024 not retry immediately. The whole system should discover the lost thing
1025 later. Keeping track with the DONE system should really be a no brainer.
1027 But there is something more: the whole filesystem is a database and the
1028 recentfiles are one possible representation of it. It's a pretty useful
1029 representation I think that's why I have implemented something around
1030 it. But for strictly local operation it has little value. For local
1031 operation we would much rather have a database. So we would enter every
1032 recentfile reading and every rsync operation and for every file the last
1033 state change and what it leads to. Then we would always ignore older
1034 records without the efforts involved with recentfiles.
1036 The database would have: path,recentepoch,rsyncedon,deletedon
1038 Oh well, not yet clear where this leads to.
1040 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1042 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1043 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1045 * statusfile, maybe only on demand, alone to have a sharp debugging
1046 tool. It is locked and all recentfiles dump themselves into it and we
1047 can build a viewer that lets us know where we stand and what's inside.
1049 * remember: only the principal recentfile needs expiration, all others
1050 shall be expired by principal if it discovers that something has move
1053 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1055 * Always check if we stringify to a higher value than in the entry
1058 * And in covered make an additional check if we would be able to see a
1059 numerical difference between the two numbers and if we can't then switch
1060 to a different, more expensive algorithm. Do not want to be caught by
1061 floating surprises. DONE
1063 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1065 * caching has several aspects here: we can cache the interval of the
1066 recentfile which only will change when the mtime of the file changes. We
1067 must re-mirror the recentfile when its ttl has expired. Does have_read
1068 tell you anything? It counts nothing at all. Only the mtime is
1069 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1070 thing the have_mirrored because it is unrelated to the mtime.
1072 * Robustness of floating point calculations! I always thought that the
1073 string calculated by the origin server for the floating representation
1074 of the epoch time is just a string. When we convert it to a number and
1075 later back to a string, the other computer might come to a different
1076 conclusion. This must not happen, we want to preserve it under any
1077 circumstances. I will have to write tests with overlong sequences that
1078 get lost in arithmetic and must see if all still works well. DONE
1080 But one fragile point remains: if one host considers a>b and the other
1081 one considers them == but no eq. To prevent this, we must probably do
1082 some extra homework. DONE
1084 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1086 * the concept of tracking DONE needs an object per recentfile that has
1087 something like these methods:
1089 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1091 covered() register() covered()
1093 The unclear thing is how we translate points in time into intervals. We
1094 could pass a reference to the current recent_events array when running
1095 we_have(xxx) and let the DONE object iterate over it such that it only
1096 has to store a list of intervals that can melt into each other. Ah, even
1097 passing the list together with a list of indexes seems feasiable.
1099 Or maybe ask for the inverted list?
1101 Whenever the complete array is covered by the interval we say we are
1102 fully covered and if the recentfile is not expired, we are uptodate.
1104 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1106 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1108 * need a way to "return" the next entry after the end of a list. When
1109 the caller says "before" or "after" we would like to know if he could
1110 cover that interval/threshold or not because this influences the effect
1111 of a newer timestamp of that recentfile. DONE with $opt{info}.
1113 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1115 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1117 loop implies tixu (?). Nope, something like --statefile decides. Per
1120 T test, I init, X including removals, U nodirtymark
1122 So we have no concept of dirtymarks, we only trust that since we are
1123 running we have observed everything steadily. But people will not let
1124 this program run forever so we must consider both startup penalty and
1125 book keeping for later runs. We keep this for later. For now we write a
1126 long running mirror that merges several intervals.
1128 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1130 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1133 * rersyncrecent, the script: default to one week. The name of the switch
1134 is --after. Other switches? --loop!
1136 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1138 * need a switch --skip-deletes (?)
1140 * need a switch --enduser that tells us that the whole tempfile
1141 discipline is not needed when there is no downstream user. (?)
1143 Without this switch we cannot have a reasonable recent.pl that just
1144 displays the recent additions. Either we accept to download everything.
1145 Or we download temporary files without the typical rsync protocol
1148 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1149 File::Temp::tempdir, this might be a win.
1151 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1153 * apropos missing: we have no push, we never know the downstream
1154 servers. People who know their downstream hosts and want to ascertain
1155 something will want additional methods we have never thought about, like
1156 update or delete a certain file.
1158 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1160 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1161 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1164 * rsync options: it's a bit of a pain that we usually need several rsync
1165 options, like compress, links, times, checksum and that there is no
1166 reasonable default except the original rsync default. I think wee can
1167 safely assume that the rsync options are shared between all recentfile
1168 instances within one recent tree.
1170 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1172 * deletes: if a delete follows an add quickly enough it may happen that
1173 a downstream mirror did not see the add at all! It seems this needs to
1174 be mentioned somewhere. The point here is that even if the downstream is
1175 never missing the principal timeframe it may encounter a "delete" that
1176 has no complimentary "add" anywhere.
1178 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1180 * I suspect the treat of metadata is incorrect during read or something.
1181 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1182 contained more than 6 hours worth of data. At 06:08 we merged into the
1183 1d file. We need to take snapshots of the 6h file over the course of an
1184 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1187 Much worse: watching the 1h file: right at the moment (at 06:35) it
1188 covers 1218867584-1219120397 which is 70 hours.
1190 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1191 08:19:44 2008, that is when I checked out last time, so it seems to be
1192 aggregating and never truncating?
1194 No, correct is: it is never truncating; but wrong is: it is aggregating.
1195 It does receive a lot of events from time to time from a larger file.
1196 Somehow a large file gets merged into the small one and because the
1197 "meta/merged" attribute is missing, nobody is paying attention. I
1198 believe that I can fix this by making sure that metadata are honoured
1199 during read. DONE and test adjusted.
1201 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1203 * grand renaming plan
1205 remotebase => remoteroot to fit well with localroot DONE
1206 local_path() => localroot seems to me should already work DONE
1207 recentfile_basename => rfilename no need to stress it has no slash DONE
1209 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1210 kern? It would anyway need a deprecation cycle because it is an
1211 important constructor.
1213 * I like the portability that Data::Serializer brings us but the price
1214 is that some day we might find out that it is slowing us a bit. We'll
1217 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1219 * should we not enter the interval of the principal (or the interval of
1220 the merging file?) in every aggregated/merged file?
1222 * we should aim at a first release and give up on thinking about
1223 sanitizing stuff and zloop. Let's just admit that a full traditional
1224 rsync is the only available sanitizer ATM. Otherwise it's complicated
1225 stuff: sanitizing on the origin server, sanitizing on the slaves,
1226 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1227 get the basics out before this becomes a major cause for mess.
1229 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1231 * On OSes not supporting symlinks we expect that RECENT.recent contains
1232 the contents of the principal recentfile. Actually this is identical on
1233 systems supporting symlinks. Simple, what follows from that is that we
1234 need to keep the serializer in the metadata because we cannot read it
1235 from the filename, doesn't it? Of course not. It's a chicken and egg
1236 problem. This leaves us with the problem to actually parse the
1237 serialized data to find out in which format it is. So who can do the 4
1238 or 5 magics we wanted to support? File::LibMagic?
1240 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1242 * remotebase and recentfile_basename are ugly names. Now that we need a
1243 word for the shortest/principal/driving recentfile too we should do
1246 localroot is good. rfile is good. local_path() is bad, local_path($path)
1247 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1250 Up to now remotebase was the string that described the remote root
1251 directory in rsync notation, like pause.perl.org::authors. And
1252 recentfile_basename was "RECENT-1h.yaml".
1254 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1256 * The test that was added in today's checkin is a good start for a test
1257 of rmirror. We should have more methods in Recent.pm: verify,
1258 addmissingfiles. We should verify the current tree, then rmirror it and
1259 then verifytree the copy. We could then add some arbitrary file and let
1260 it be discovered by addmissingfiles, then rmirror again and then
1261 verifytree the copy again.
1263 Then we could start stealing from csync2 sqlite database [no port to
1264 OSX!] and fill a local DB. And methods to compare the database with the
1265 recentfiles. Our strength is that in principle we could maintain state
1266 with a single float. We have synced up to 1234567890.123456. If the Z
1267 file does not add new files all we have to do is mirror the new ones and
1270 This makes it clear that we should extend current protocol and declare
1271 that we cheat when we add files too late, just to help the other end
1272 keeping track. Ah yes, that's what was meant when zloop was mentioned
1275 Maybe need to revisit File::Mirror to help me with this task.
1277 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1279 * There must be an allow-me-to-truncate flag in every recentfile.
1280 Without it one could construct a sequence of updates winning the locking
1281 battle against the aggregator. Only if an aggregator has managed to
1282 merge data over to the next level, truncating can be allowed. DONE with
1285 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1287 * We should probably guarantee that no duplicates enter the aggregator
1290 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1292 * To get merge operation faster would need a good benchmark test. What
1293 02 spits out isn't reliable enough and is dominated by many other
1296 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1297 Author: Andreas J. Koenig <k@k75.(none)>
1298 Date: Sat Aug 2 07:58:04 2008 +0200
1302 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1303 Author: Andreas J. Koenig <k@k75.(none)>
1304 Date: Sat Aug 2 11:40:29 2008 +0200
1306 there was a lot of trying but the effect is hardly measurable with
1309 * overhead of connecting seems high. When setting
1310 max_files_per_connection to 1 we see that.
1312 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1314 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1316 25138.2215600014/3600 = 6.98283932222261
1318 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1319 be a bug in the aggregator. FIXED
1321 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1323 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1324 we have it, can we mirror it, mirror it!
1326 I fear this needs a new class which might be called
1327 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1328 and treat them as an entity. I realize that a single recentfile may be
1329 sufficient for certain tasks and that it is handy for the low level
1330 programmer but it is not nice to use. If there is a delete in the 1h
1331 file then the 6h file still contains it. Seekers of the best information
1332 need to combine at least some of the recentfiles most of the time.
1334 There is the place for the Z loop!
1336 But the combination is something to collect in a database, isn't it. Did
1337 csync2 just harrumph?
1339 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1341 * it just occurred to me that hosts in the same mirroring pool could
1342 help out each other even without rewriting the recentfile. Just fetch
1343 the stuff to mirror from several places, bingo. But that's something
1344 that should rather live in a separate package or in rsync directly.
1346 * cronjobs are unsuited because with ntp they would all come at the full
1347 minute and disturb each other. Besides that I'd hate to have a backbone
1348 with more than a few seconds latency.
1350 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1352 * a second rsync server with access control for PAUSE. Port? 873 is the
1353 standard port, let's take 8873.
1355 * if there were a filesystem based on this, it would have a slow access
1356 to inexistent files. It would probably provide wrong readdir (only based
1357 on current content) or also a slow one (based on a recentfile written
1358 after the call). But it would provide fast access to existing files. Or
1359 one would deliberately allow slightly blurred answers based on some
1360 sqlite reflection of the recentfiles.
1362 * todo: write a variant of mirror() that combines two or more
1363 recentfiles and treats them like one
1365 * todo: signal handler to remove the tempfile
1367 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1369 * now that we have the symlink I forgot how it should be used in
1372 * the z loop: add missing files to Z file. Just append them (instead of
1373 prepending). So one guy prepends something from the Y file from time to
1374 time and another guy appends something rather frequently. Collecting
1375 pond. When Y merges into Z, things get epoch and the collecting pond
1376 gets smaller. What exactly are "missing files"?
1378 take note of current epoch of the alpha file, let's call it the
1381 find all files on disk
1383 remove all files registered in the recentworld up to recent-ts
1385 remove all files that have been deleted after recent-ts according to
1388 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1390 * rersyncrecent might be a cronjob with a (locked) state file which
1391 contains things like after and maybe last z sync or such?
1393 rrr-mirror might be an alternative name but how would we justify the
1394 three Rs when there is no Re-Rsync-Recent?
1396 With the --loop parameter it is an endless loop, without it is no loop.
1397 At least this is simple.
1399 * todo: new accssor z-interval specifies how often the Z file is updated
1400 against the filesystem. We probably want no epoch stamp on these
1401 entries. And we want to be able to filter the entries (e.g. no
1402 by-modules and by-category tree)
1404 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1406 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1407 into Z. We do not need the other files filled up with filesystem
1410 * need interface to query for a file in order to NOT call update on
1411 PAUSE a second time within a short time.
1413 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1415 * recommended update interval? Makes no sense, is different for
1422 change-log-default-name: "Todo"