| blob | 7cd4f4166337a08f494d6a4b9dcf6c4d987b495f |
1 2010-10-26 Andreas J. Koenig <andk@cpan.org>
3 * forking slave is done, needs more testing.
5 2010-07-06 Andreas J. Koenig <andk@cpan.org>
7 * The two next most important goals are to get a useable rrr-server and
8 a forking slave implementation that uses virtually no memory after the
9 first pass through. Does it need the status file?
11 * rrr-init now being fast we can now remove the alpha alerts. Done.
13 2010-06-16 Andreas J. Koenig <andk@cpan.org>
15 * rrr-init seems to be sssslllloooowwww. At 3:49 I started it for the
16 video tree. After twelve minutes the file has reached a size of 500k. It
17 is locking and rewriting the YAML after every file. Wasn't there an
18 unlocked variant? For "init" one would not expect locked correctness and
19 a rewrite after every item. Given that this tends to become slower with
20 every file it is really unacceptable.
22 Reminds me of rrr-dirtyupdate (the former name of rrr-update). It seems
23 the code was not dirty enough and that was the reason why it was so
24 slow. And rrr-update does not have a --dirty switch, so it looks like we
25 have no way to force dirty and fast operation yet but we need it for
26 init.
28 Irritating the bootstrapping race: the order {(1) fill initial
29 array, (2) start server} leaves the need for an fsck to cover the time
30 between 1 and 2. Given that regular fsck is needed in any case the price
31 is not high but still doesn't give a warm feeling. Maybe starting the
32 server will by itself do an fsck?
34 2010-06-15 Andreas J. Koenig <andk@cpan.org>
36 * remove the wild alpha alerts and say, it is alpha but parts of it are
37 already working very well or so.
39 2009-12-10 Andreas J. Koenig <andk@cpan.org>
41 * Prominent request for statusfile to safe memory consumption. I'd like
42 to work backwards: get a verification program that verifies how well in
43 sync two nodes are without using frmr itself. Pretend that the forking
44 agent already works and can be interrupted with ^C. Then let it die at
45 random and restart at random and see the memory consumption and the
46 failure rate over time. Compare it with the curve we have now.
48 Of course we also invented the status file as a sharp debugging tool.
49 I'd like to see it in action. See also: F:R:M:Recent::thaw().
51 2009-06-22 Andreas J. Koenig <andk@cpan.org>
53 * Todo: rrr-init to generate some index files so that rrr-server can be
54 started.
56 2009-05-02 Andreas J. Koenig <andk@cpan.org>
58 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
59 so we can use it for normal update with a single file. And add a cronjob
60 that runs rrr-fsck so we get a reminder when something is not working
61 as it should.
63 2009-04-27 Andreas J. Koenig <andk@cpan.org>
65 * Todo: I still see more link_stat errors than expected. I come back to
66 the former suspicion that we do not do necessary bookkeeping on deleted
67 files so that when we reach the "new" event on a meanwhile deleted file
68 we try to mirror it although we could know better. It's just noise, no
69 harm but noise is irritating.
71 * Todo: more experimenting with runstatusfile to get it official
73 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
74 parse the error.
76 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
78 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
80 And that ignore_link_stat_error is propagated in sparse_clone.
82 DONE
84 For the record: the bug was that ignore_link_stat_error was lost during
85 sparse_clone. And then the second bug was that it did not default to
86 true. Both are now fixed.
88 2009-04-25 Andreas J. Koenig <andk@cpan.org>
90 * Todo: fill rrr program with life
92 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
94 * Minor bug: I see complaints about files not existing like
95 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
96 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
98 In the RECENT files these are delete events but of course also new
99 events in older recent files.
101 I think we do not properly remember deletes when we skip-deletes and so
102 do not filter them out when we later see the "new" event.
104 Yes, but that what ignore_link_stat_error is for, we know about this
105 race condition. We must make ignore_link_stat_error default to true.
107 DONE
109 * bug in Done.pm failing to merge two fields into one when a third is
110 present. FIXED and accompanied with new test.
112 2009-04-24 Andreas J. Koenig <andk@cpan.org>
114 * Todo: keep the index clean and run some kind of nondestructive fsck 4
115 times a day.
117 * who is our backbone?
119 http://cpan.cpantesters.org/authors/02STAMP # barbie
120 http://cpan.solfo.com/authors/02STAMP # abh
121 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
123 2009-04-17 Andreas J. Koenig <andk@cpan.org>
125 * the changelog helper I included in the Makefile release memo does not
126 show tags and is much less useful than I thought. gitk is probably much
127 more convenient.
129 2009-04-16 Andreas J. Koenig <andk@cpan.org>
131 * ABH writes:
133 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
134 #### Ydr_.yaml ... DONE
135 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
136 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
137 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
138 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
139 #### gone already at cpan-pause.pl line 0
141 Running without skip-deletes now but cannot reproduce.
143 2009-04-15 Andreas J. Koenig <andk@cpan.org>
145 * consider state directory such we can restart after a ^C or come again
146 from a cronjob.
148 Reuse _runstatusfile: write atomically, see how we can use the dump to
149 start running again. Do not leak and remove globs before writing. Do not
150 only dump one rf, dump the whole complex.
152 * consider the effect of ^C. Is it safe?
154 * want to use Perl::Version to up the versions on every release. Only if
155 things have changed, of course. Does this actually work?
157 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
158 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
160 Must do the bookkeeping myself when this should happen, maybe based on git?
162 I think I prefer the same version number for all pm files. But one also
163 needs a routine for setting them. Ahh:
165 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
167 NEEDSMOREWORK
169 * why indexer wrong?
171 cpan[2]> m /Mirror::Recent/
172 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
173 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
174 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
175 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
176 4 items found
178 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
180 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
182 I just tried MM 6.5102 and it again produces only a provides for Recent,
183 not for the other modules.
185 My own fault. FIXED in the Makefile.PL.
187 2009-04-13 Andreas J. Koenig <andk@cpan.org>
189 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
191 * is it true (as stated at
192 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
193 that rsync can lead to errors when upstream changes before client sync
194 has completed? Can we see the error when we run this on upstream:
196 % perl -e '
197 use Time::HiRes qw(time);
198 while (){
199 open my $fh, ">", "changingfile.txt.new" or die;
200 print $fh (time."\n") x 1000000;
201 close $fh;
202 rename "changingfile.txt.new", "changingfile.txt" or die;
203 }
204 '
206 And on the receiving end:
208 % while true; do
209 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
210 sleep 1;
211 done
213 I cannot get it to fail. Apparently it is sufficient when upstream
214 always writes atomically (which of course is mandatory)?
216 SENT to mailing list Apr 28.
218 * from ABH:
220 > https://fedorahosted.org/InstantMirror/
221 > https://www.redhat.com/mailman/listinfo/instantmirror-list
222 > irc.freenode.net channel #instantmirror
224 A cool name for a project. Inspires me to write a few sentences about
225 bittorrent's role in the grand picture.
227 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
228 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
230 2009-04-12 Andreas J. Koenig <andk@cpan.org>
232 * Interesting last minute bug during real download testing: the output
233 isn't as pretty anymore, it seems that more work is being done than
234 needed.
236 Introducing a highlevel _alluptodate method helped a lot in debugging
237 this.
239 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
240 different on 1h file than on the other files. This lead to frequent
241 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
242 issue.
244 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
246 * Todo: try if we can be faster with a native float package. I'm really
247 glad to have a machine exposing perl floating point bugs, it forced me
248 to retract this item from the changes file: "several speedups in the
249 fakebigfloat code" before I get cpan testers fails.
251 2009-04-11 Andreas J. Koenig <andk@cpan.org>
253 * Bug: when the dirtymark changes upstream, then the downstream server
254 notices it immediately and mirrors the "1h" file but then leaves the
255 rmirror loop and continues with "6h" on the next round through. This
256 behaviour goes on until it reaches the Z file. Because we eat the Z file
257 piecemeal we leave the loop after a while but after that everybody seems
258 to have forgotten that there is some work left to be done. Bad, bad,
259 bad.
261 ETOOCONFUSING better name?
263 Recentfile::get_remote_recentfile_as_tempfile OK
264 Recentfile::resolve_recentfilename split_rfilename DONE
266 Recent::_principal_recentfile_object _principal_recentfile_fromremote
267 Recent::principal_recentfile OK
268 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
270 some void, some not, some not void but called in void context.
272 2009-04-10 Andreas J. Koenig <andk@cpan.org>
274 * Bug? When the dirtymark changes then the second tier hosts quickly
275 reset their DONE state and restart mirroring. But then they mirror
276 potentially outdated index files with inconsistent dirtymark because the
277 first tier box may be a bit behind on the large recentfiles.
279 This means we need a brake when iterating over the recentfiles that
280 refuses to use a recentfile with the wrong dirtymark.
282 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
284 * install the bin/* files and talk about rrr-overview somewhere.
286 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
287 created outside the target tree.
289 * Barbie asks for better logging capabilities
291 2009-03-29 Andreas J. Koenig <andk@cpan.org>
293 * known bug: cannot configure the Recentfile objects to keep delete
294 events. How to call the accessor for that? keep_delete_objects_forever?
296 Sounds acceptable. FIXED (but untested).
298 * known bug: keeps temporary index files lying around.
300 2009-03-24 Andreas J. Koenig <andk@cpan.org>
302 * Some equivalent for
304 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
306 ? DONE
308 2009-03-22 Andreas J. Koenig <andk@cpan.org>
310 * k81 is again client of k75 with different parameters.
312 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
313 about it all the time and nobody reacts accordingly; (2) lots of temp
314 files get created and not removed; culprit the new call to
315 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
316 caller has to remove the tempfile after use.
318 Need the drawing board.
320 PARTLY FIXED: retracting the idea to call
321 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
322 and unseed such that uptodate now means "we have mirrored this
323 recentfile's bunch of files" and seeded means "somebody believes the
324 index file for this rf needs be refreshed". This seems to work now.
326 Still collecting temp files that nobody cares to depollute.
328 2009-03-21 Andreas J. Koenig <andk@cpan.org>
330 * Bug?: should it be harder than it is atm to set the timestamp to the
331 future? YES, FIXED
333 * bug with native integers:
335 -
336 epoch: 997872011
337 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
338 type: new
339 -
340 epoch: 1195248431
341 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
342 type: new
344 Native integer broke when native math was turned off. FIXED
346 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
347 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
348 in this area.
350 Yes, records out of order:
352 447003 -
353 447004 epoch: 995885533
354 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
355 447006 type: new
356 447007 -
357 447008 epoch: 995890358
358 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
359 447010 type: new
360 447011 -
361 447012 epoch: 995892221
362 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
363 447014 type: new
365 FIXED with sanity check and later with the integer fix.
367 * Bug: want the index files in a .recent directory
369 * Bug: lots of dot files are not deleted in time
371 * possible test case: can a delete change the timestamp? This would
372 probably break the order of events.
374 2009-03-20 Andreas J. Koenig <andk@cpan.org>
376 * 1233701831.34486 what's so special about this number/string? It
377 belongs to
378 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
379 atm lives in Y,Q, and Z.
381 It is the first entry after id/--skip-locking which has timestamp
382 1234164228.11325 which represents a file that doesn't exist anymore.
384 * another bug:
386 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
387 _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
388 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
389 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
390 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
391 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
392 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
393 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
394 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
395 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
396 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
397 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
398 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
400 FIXED, it was the "--skip-locking" file where manual intervention was participating
402 * bug on the mirroring slave: when the dirtymark gets increased we
403 probably do not reset the done intervals. The mirrorer stays within
404 tight bounds where it tries to sync with upstream and never seems to
405 finish. In the debugging state file I see lots of identical intervals
406 that do not get collapsed. When I restart the mirrorer it dies with:
408 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
409 _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
410 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
411 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
412 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
413 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
414 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
415 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
416 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
417 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
419 and debugging stands at
421 At: before
422 Brfinterval: Z
423 Ci: 227
424 Dre+1:
425 epoch: 1237400802.5789
426 path: id/J/JO/JOHND/CHECKSUMS
427 type: new
428 Dre-0:
429 epoch: 1237400807.97514
430 path: id/M/MI/MIYAGAWA/CHECKSUMS
431 type: new
432 Dre-1:
433 epoch: 1237400817.94363
434 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
435 type: new
436 Eintervals:
437 -
438 - 900644040
439 - 900644040
440 - []
442 and it is reproducable.
444 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
445 pause has a fresh one.
447 FIXED, it was the third anded term in each of the ifs in the IV block in
448 _register_one: with that we make sure that we do not stamp on valuable
449 interval data.
451 2009-03-17 Andreas J. Koenig <andk@cpan.org>
453 * done: verified the existence of the floating point bug in bleadperl
454 and verified that switching from YAML::Syck to YAML::XS does not resolve
455 it.
457 BTW, the switch was doable with
459 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
461 and should be considered as a separate TODO
463 * todo: integrate a dirty update with two aggregate calls before
464 unlocking for frictionless dirtying DONE
466 * todo: start the second rsync daemon on pause DONE
468 * todo: move index files to .recent: this cannot simply be done by
469 setting filenameroot to .recent/RECENT. Other parts of the modules rely
470 on the fact that dirname(recentfile) is the root of the mirrored tree.
472 2009-03-16 Andreas J. Koenig <andk@cpan.org>
474 * What was the resolution of the mirror.pl delete hook bug? Do we call
475 the delete hook when pause removes a file from MUIR?
477 * Today on pause: Updating 2a13fba..29f284d and installing it for
478 /usr/local/perl-5.10.0{,-RC2}
480 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
481 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
483 2009-03-15 Andreas J. Koenig <andk@cpan.org>
485 * currently recent_events has the side effect of setting dirtymark
486 because it forces a read on the file. That should be transparent, so
487 that the dirtymark call always forces a cache-able(?) read.
489 * The bug below is -- after a lot of trying -- not reproducible on a
490 small script, only in the large test script. The closest to the output
491 below was:
493 #!perl
494 use strict;
495 use Devel::Peek;
496 my $x = "01237123229.8814";
497 my($l,$r);
498 for ($l,$r) {
499 $_ = "x"x34;
500 }
501 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
502 $r = int $r;
503 $l = "1237123231.22458";
504 $r = "1237123231.22458";
505 1 if $l/1.1;
506 Devel::Peek::Dump $l;
507 Devel::Peek::Dump $r;
508 Devel::Peek::Dump $x = $l <=> $r;
509 die "BROKE" if $x;
510 __END__
512 The checked in state at c404a85 fails the test with my
513 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
514 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
515 in the test script always consistent.
517 * Todo: write a test that inserts a second dirty file with an already
518 existing timestamp. DONE
520 * Bug in perl 5.10 on my 64bit box:
522 DB<98> Devel::Peek::Dump $l
523 SV = PVMG(0x19e0450) at 0x142a550
524 REFCNT = 2
525 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
526 IV = 0
527 NV = 1237123231.22458
528 PV = 0x194ce70 "1237123231.22458"\0
529 CUR = 16
530 LEN = 40
532 DB<99> Devel::Peek::Dump $r
533 SV = PVMG(0x19e0240) at 0x142a3e8
534 REFCNT = 2
535 FLAGS = (PADMY,POK,pPOK)
536 IV = 1237123229
537 NV = 1237123229.8814
538 PV = 0x19ff900 "1237123231.22458"\0
539 CUR = 16
540 LEN = 40
542 DB<100> Devel::Peek::Dump $l <=> $r
543 SV = IV(0x19ea6e8) at 0x19ea6f0
544 REFCNT = 1
545 FLAGS = (PADTMP,IOK,pIOK)
546 IV = -1
548 DB<101> Devel::Peek::Dump $l
549 SV = PVMG(0x19e0450) at 0x142a550
550 REFCNT = 2
551 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
552 IV = 0
553 NV = 1237123231.22458
554 PV = 0x194ce70 "1237123231.22458"\0
555 CUR = 16
556 LEN = 40
558 DB<102> Devel::Peek::Dump $r
559 SV = PVMG(0x19e0240) at 0x142a3e8
560 REFCNT = 2
561 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
562 IV = 1237123231
563 NV = 1237123231.22458
564 PV = 0x19ff900 "1237123231.22458"\0
565 CUR = 16
566 LEN = 40
568 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
570 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
572 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
574 * $obj->merge ($other) needs to learn about equal epoch which may happen
575 since dirty_epoch intruded.
577 * Wontfix anytime soon: I think we currently do not support mkdir. Only
578 files!
580 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
582 * Todo: continue working on update(...,$dirty_epoch). It must be
583 followed by a fast_aggregate! DONE
585 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
587 * maybe we need a closest_entry or fitting_interval or something like
588 that. We want to merge an event into the middle of some recentfile.
589 First we do not know which file, then we do not know where to lock,
590 where to enter the new item, when and where to correct the dirtymark.
592 So my thought is we should first find which file.
594 Another part of my brain answers: what would happen if we would enter
595 the new file into the smallest file just like an ordinary new event,
596 just as an old event?
598 (1) we would write a duplicate timestamp? No, this would be easy to
599 avoid
601 (2) we would make the file large quickly? Yes, but so what? We are
602 changing the dirtymark, so are willing to disturb the downstream hosts.
604 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
606 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
608 leak!
610 https://rt.cpan.org/Ticket/Display.html?id=41199
612 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
614 Strangest output being HAKANARDO who managed to upload
616 Here is a better oneliner that includes also the first line of each
617 finding:
619 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
621 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
623 * memory leak in the syncher? It currently weighs 100M.
625 Update 2008-11-02:
627 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
630 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
632 * lookup by epoch and by path and use this ability on the pause to never
633 again register a file twice that doesn't need it. Let's call it
634 contains().
636 * after the dirtymark is done: fill up recentfiles with fake (historic)
637 entries; fill up with individual corrections; algorithm maybe to be done
638 with bigfloat so that we can always place something in the middle
639 between two entries. Before we must switch to bigfloat we could try to
640 use Data::Float::nextup to get the.
642 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
643 directly.
645 * dirtymark now lives in Recentfile, needs to be used in rmirror.
647 * find out why the downloader died after a couple of hours without a net
648 connection. Write a test that survives the not-existence of the other
649 end forever.
651 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
653 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
654 it makes less sense with tight coupling for secondary files. But asking
655 the server what the current epoch is might be cheaper on HTTP than on
656 rsync. (Needs to be evaluated)
658 * remove the 0.00 from the verbose overview in the Merged column in the
659 Z row. DONE
661 * write tests that expose the problems of the last few days: cascading
662 client/server roles, tight coupling for secondary RFs, deletes after
663 copies.
665 * Some day we might want to have policy options for the slave:
666 tight/loose/no coupling with upstream for secondary RFs. tight is what
667 we have now. loose would wait until a gap occurs that can be closed.
669 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
671 * revisit all $rfs->[$i+1] places if they now make sense still
673 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
675 * another bug is the fact that the mirror command deletes files before
676 it unhides the index file, thus confusing downstream slaves. We must not
677 delete before unhiding and must delete after unhiding. FIXED.
679 * new complication about the slave that is playing a server role.
680 Currently we mirror from newest to oldest with a hidden temporary file
681 as index. And when one file is finished, we unhide the index file.
682 Imagine the cascading server/slave is dead for a day. It then starts
683 mirroring again with the freshest thing and unhides the freshest index
684 file when it has worked through it. In that moment it exposes a time
685 hole. Because it now works on the second recentfile which is still
686 hidden.
688 We currently do nothing special to converge after such a drop out. At
689 least not intentionally and robustly and thought through.
691 The algorithm we use to seed the next file needs quite a lot of more
692 robustness than it currently has. Something to do with looking at the
693 merged element of the next rf and when it has dropped off, we seed
694 immediately. And if it ramains dropped off, we seed again, of course.
696 Nope, looking from smaller to larger RFS we look at the merged element
697 of this RF and at the minmax/max element of the next RF. If that
698 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
699 seeding it.
701 And we need a public accessor seed and unseed or seeded. But not the mix
702 of public and private stuff that then is used behind the back.
704 And then the secondary* stuff must go.
706 And we must understand what the impact is on the DONE system. Can it go
707 unnoticed that there was a hole? And could the DONE system have decided
708 the hole is covered? This should be testable with three directories where
709 the middle stops working for a while. Done->merge is suspicious, we must
710 stop it from merging non-conflatable neighbors due to broken continuity.
712 FIXED
714 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
716 * Slaven suggests to have the current epoch or the whole current
717 recentfile available from the HTTP server and take it away with
718 keepalive. This direction goes the granularity down to subseconds.
720 We might want to rewrite everything to factor out transport and allow
721 the whole thing to run via HTTP.
723 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
725 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
726 symlink?
728 * smoker on k81 fetching from k75 to verify cascading works. See
729 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
731 * maybe the loop should wait for CHECKSUMS file after every upload. And
732 CPAN.pm needs to deal with timestamps in the future.
734 * do not forget the dirtymark!
736 Text: have a new flag on recentfiles with the meaning: if this
737 changes, you're required to run a full rsync over all the files. The
738 reason why we set it would probably be: some foul happened. we injected
739 files in arbitrary places or didn't inject them although they changed.
740 The content of the flag? Timestamp? The relation between the
741 recentfiles would have to be inheritance from the principal, because any
742 out of band changes would soon later propagate to the next recentfile.
744 By upping the flag often one can easily ruin the slaves.
746 last out of band change? dirtymark?
748 Anyway, this implies that we read a potentially existing recentfile
749 before we write one.
751 And it implies that we have an eventloop that keeps us busy in 2-3
752 cycles, one for current stuff (tight loop) and one for the recentfiles
753 (cascade when principal has changed), one for the old stuff after a
754 dirtymark change.
756 And it implies that the out-of-band change in any of the recentfiles
757 must have a lock on the principal file and there is the place to set the
758 dirtymark.
760 * start a FAQ, especially quick start guide questions. Also to aid those
761 problematic areas where we have no good solution, like the "links"
762 option to rsync.
764 * wish feedback when we are slow.
766 * reduce mccabe
768 * Remove a few DEBUG statements.
770 * The multiple-rrr way of doing things needs a new option to rmirror,
771 like piecemeal or so. Not urgent because after the first pass through,
772 things run smoothely. It's only ugly during the first pass.
774 * I have the suspicion that the code is broken that decides if the
775 neighboring RF needs to be seeded. I fear when too much time has gone
776 between two calls (in our case more than one hour), it would not seed
777 the neighbor. Of course this will never be noticed, so we need a good
778 test for it.
780 * local/localroot confusion: I currently pass both options but one must
781 do.
783 * accounts for early birds on PAUSE rsync daemon.
785 * hardcoded 20 seconds
787 * who mirrors the index? DOING now.
789 * which CPAN mirrors offer rsync?
791 * visit all XXX, visit all _float places
793 * rename the pathdb stuff, it's too confusing. No idea how.
795 * rrr-inotify, backpan, rrr-register
797 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
799 * current bugs: the pathdb seems to get no reset, the seeding of the
800 secondaryttl stuff seems not to have an effect. Have helped myself with
801 a rand(10), need to fix this back. So not checked in. Does the rand
802 thing even help?
804 The rand thing helps. The secondaryttl stuff was in the wrong line,
805 fixed now.
807 The pathdb stuff was because I called either _pathdb or __pathdb on the
808 wrong object. FIXED now.
810 * It's not so beautiful if we never fetch the recentfiles that are not
811 the principal, even if this is correct behaviour. We really do not need
812 them after we have fetched the whole content.
814 OK, we want a switch for that: secondaryttl DONE
816 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
818 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
819 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
820 it lets 40 follow. FIXED
822 * See that the long running process really only updates the principal
823 file unless it has missed a timespan during which something happened. If
824 nothing happened, it must notice even when it misses the timespan. DONE
826 * we must throw away the pathdb when we have reached the end of Z. From
827 that moment we can have a very small pathdb because the only reason for
828 a pathdb is that we know to ignore old records in old files. We won't
829 need this pathdb again before the next full pass over the data is
830 necessary and then we will rebuild it as we go along. DONE
832 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
834 * I think, Done::register_one is doing wrong in that it does not
835 conflate neighboring pieces. The covered() method cannot do this because
836 it has no recent_events array at hand. But register_one has it and could
837 do it and for some reason misses to do it (sometimes).
839 This means that the three tests I just wrote can probably not survive
840 because they test with an already broken Done structure.
842 The art now is to detect how it happens, then to reproduce, then write a
843 test, then fix it.
845 So from the logfile this is what happens: we have a good interval with
846 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
847 top of it. Locally we now mirror F2 and open a new done interval for it.
848 Then we mirror F1 but this time with the timestamp T1b. And when we then
849 try to close the gap, we do not find T1 but instead something older. We
850 should gladly accept this older piece and this would fix this bug.
852 FIXED
854 * bug to fix: when the 1h file changes while rmirror is running, we do
855 correctly sync the new files but never switch to the 6h file but rather
856 stay in a rather quick loop that fetches the 1h file again and again.
858 Is it possible that we initialize a new object? Or does
859 get_remote_recentfile_as_tempfile overwrite something in myself?
861 Want a new option: _runstatusfile => $file which frequently dumps the
862 state of all recentfiles to a file.
864 FIXED
866 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
868 * Todo: now teach update to verify the timestamp is about to write
869 against the previous and use _increase_a_bit if it doesn't comply with
870 strict monotony. DONE
872 * The problem of rounding. So far perl's default precision was
873 sufficient. One day it won't be. FakeFloat has an easy job when it is
874 only reading and other machines have written correctly. But when we want
875 to write a floating point number that is a bit larger than the other
876 one, then we need our own idea of precision.
878 Slaven said: just append a "1". This might be going towards the end of
879 usability too quickly. I'd like something that actually uses the decimal
880 system. Well, appending a 1 also does this but...
882 E.g. we have 1.0. nextup on this architecture is starting with
883 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
884 taken the 1.0000000000000003 and the next user comes and the time tells
885 him 1.0 again. He has to beat my number without stepping over the
886 nextup. This is much less space than I had when I chose 1,2,3.
888 What is also irritating is that nextup is architecture dependent. The
889 128 bit guy must choose very long numbers to fit in between whereas the
890 other one with 16 bit uses larger steps. But then the algorithm is the
891 same for both, so that would be a nice thing.
893 I see two situation where we need this. One is when Time::HiRes returns
894 us a value that is <= the last entry in our recentfile. In this case
895 (let's call it the end-case) we must fill the region between that number
896 and the next higher native floating point number. The other is when we
897 inject an old file into an old recentfile (we would then also set a new
898 dirtymark). We find the integer value already taken and need a slightly
899 different one (let's call it the middle-case). The difference between
900 the two situations is that the next user will want to find something
901 higher than my number in the end-case and something lower than my number
902 in the middle case.
904 So I suggest we give the function both a value and an upper bound and it
905 calculates us a primitive middle. The upper bound in the middle-case is
906 the next integer. The upper bound on the end-case is the nextup floating
907 point number. But the latter poses another problem: if we have occupied
908 the middle m between x and nextup(x), then the nextup(m) will probably
909 not be the same as nextup(x) because some rounding will take place
910 before the nextup is calculated and when the rounding reaches the
911 nextup(x), we will end up at nextup(nextup(x)).
913 So we really need to consider the nextup and the nextdown from there and
914 then the middle and that's the number we may approach asymptotically.
915 Ugly. But DONE.
917 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
919 * consider deprecating the use of RECENT.recent as a symlink. It turns
920 out to need extra hoops with the rsync options and just isn't worth it.
921 Or maybe these extra hoops are needed anyway for the rest of the tree?
922 Nope, can't be the case because not all filesystems support symlinks.
924 But before doing the large step, I'll deprecate the call of
925 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
926 only introduced to resolve RECENT.recent and complicates the routine far
927 beyond what it deserves.
929 DONE. Won't deprecate RECENT.recent, just moved its handling to the
930 supervisor.
932 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
934 * I think it's a bug that the rsync_option links must be set to true in
935 order to support RECENT.recent and that nobody cares to set it
936 automatically. Similar for ignore_link_stat_errors.
938 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
940 * Todo: find all todos together and make a plan what is missing for a
941 release.
943 - verifytree or something like that. fsck maybe.
945 - rersyncrecent, the script itself? What it do?
947 - a way to only mirror the recentfiles without mirroring the whole
948 remote system such that people can decide to mirror only partially see
949 also 2008-08-30. .shadow-xxx directory? this also needed for a
950 filesystem that is still incomplete and might need the mirrorfiles for
951 lookup(?)
953 - long living objects that mirror again and again. Inject something
954 into ta, see how it goes over to tb.
956 - how do we continue filling up the DONE system when we use an object
957 for the second time? "fully covered" and "uptodate" or new terminology.
959 - overview called on the wrong file should be understandable
961 - the meta data field that must change when we fake something up so that
962 the downstream people know they have to re-fetch everything.
964 - how tolerant are we against missing files upstream? how do we keep
965 track? there are legitimate cases where we did read upstream index right
966 before a file got deleted there and then find that file as new and want
967 it. There are other cases that are not self healing and must be tracked
968 and bugreported.
970 - how, exactly, do we have to deal with deletes? With rsync errors?
972 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
973 authors) failed: No such file or directory (2)
975 The file above is a delete in 1h and a new in file 1M and the
976 delete in the locally running rmirror did not get propagated to the 1M
977 object. Bug. And the consequence is a standstill.
979 It seems that a slave that works with a file below the principal needs
980 to merge things all the way up to get rid of later deletes. Or keep
981 track of all deletes and skip them later. So we need a trackdeletes.pm
982 similar to the done.pm?
984 see also 2008-08-20 about spurious deletes that really have no add
985 counterpart and yet they are not wrong.
987 - consider the effect when resyncing the recentfile takes longer than
988 the time per loop. Then we never rsync any file. We need to diagnose
989 that and force an increase of that loop time. But when we later are fast
990 enough again because the net has recovered, then we need to switch back
991 to original parameters. ERm, no, it's enough to keep syncing at least
992 one file before refetching an index file.
994 - remember to verify that no temp files are left lying around and the
995 signal handler
997 - status file for not long running jobs that want to track upstream with
998 a, say, cronjob.
1000 - revisit all XXX _float areas and study Sub::Exporter DONE
1002 - persistent DB even though we just said we do not need it. Just for
1003 extended capabilities and time savings when, for example, upstream
1004 announces a reset and we get new recentfiles and could then limit
1005 ourselves to a subset of files (those that have a changed epoch) in a
1006 first pass and would only then do the loop to verify the rest. Or
1007 something.
1009 * Todo: aggregate files should know their feed and finding the principal
1010 should be done stepwise. (?)
1012 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
1013 left locks around. DONE
1015 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1017 * maybe extend the _overview so that it always says if and where the
1018 last file is in the next file and where the next event in the next rf
1019 would lie. No, don't like this anymore. REJECT
1021 * take the two new redundant tests out again, only the third must
1022 survive. DONE
1024 * Todo: add a sanity check if the merged structure is really pointing to
1025 a different rf and that this different rf is larger. DONE
1027 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1029 * now test, if they are overlapping. And test if there is a file in the
1030 next rf that would fit into this rf's interval.
1032 1h 1222324012.8474 1222322541.7963 0.4086
1033 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
1034 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
1035 1W 1222313218.3626 1221708477.5829 167.9835
1037 I suspect that somebody writes a merged timestamp without having merged
1038 and then somebody else relies on it.
1040 If aggregate is running, the intervals must not be extravagated, if it
1041 is not running, there must not be bounds, the total number of events in
1042 the system must be counted and must be controlled throughout the tests.
1043 That the test required the additional update was probably nonsense,
1044 because aggregate can cut pieces too. FIXED & DONE
1046 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1048 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1049 changed. FIXED
1051 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1053 * Most apparent bug at the moment is that the recentfiles are fetched
1054 too often. Only the principal should be fetched and if it has not
1055 changed, the others should not be refetched. ATM I must admit that I'm
1056 happy that we refetch more often than needed because I can more easily
1057 fix bugs while the thing is running.
1059 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1060 done but the done system does not know about it. The reason for the
1061 failure is not known and we never reach the status uptodate because of
1062 this. We must get over it.
1064 Later it turns out that the origin server had a bug somewhere.
1065 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1066 array of the recentfile one position above. The bug was my own.
1068 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1070 * There is the race condition where the server does a delete and the
1071 slave does not yet know and then tries to download it because he sees
1072 the new. So for this time window we must be more tolerant against
1073 failure. If we cannot download a file, we should just skip it and should
1074 not retry immediately. The whole system should discover the lost thing
1075 later. Keeping track with the DONE system should really be a no brainer.
1077 But there is something more: the whole filesystem is a database and the
1078 recentfiles are one possible representation of it. It's a pretty useful
1079 representation I think that's why I have implemented something around
1080 it. But for strictly local operation it has little value. For local
1081 operation we would much rather have a database. So we would enter every
1082 recentfile reading and every rsync operation and for every file the last
1083 state change and what it leads to. Then we would always ignore older
1084 records without the efforts involved with recentfiles.
1086 The database would have: path,recentepoch,rsyncedon,deletedon
1088 Oh well, not yet clear where this leads to.
1090 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1092 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1093 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1095 * statusfile, maybe only on demand, alone to have a sharp debugging
1096 tool. It is locked and all recentfiles dump themselves into it and we
1097 can build a viewer that lets us know where we stand and what's inside.
1099 * remember: only the principal recentfile needs expiration, all others
1100 shall be expired by principal if it discovers that something has move
1101 upstream.
1103 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1105 * Always check if we stringify to a higher value than in the entry
1106 before. DONE
1108 * And in covered make an additional check if we would be able to see a
1109 numerical difference between the two numbers and if we can't then switch
1110 to a different, more expensive algorithm. Do not want to be caught by
1111 floating surprises. DONE
1113 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1115 * caching has several aspects here: we can cache the interval of the
1116 recentfile which only will change when the mtime of the file changes. We
1117 must re-mirror the recentfile when its ttl has expired. Does have_read
1118 tell you anything? It counts nothing at all. Only the mtime is
1119 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1120 thing the have_mirrored because it is unrelated to the mtime.
1122 * Robustness of floating point calculations! I always thought that the
1123 string calculated by the origin server for the floating representation
1124 of the epoch time is just a string. When we convert it to a number and
1125 later back to a string, the other computer might come to a different
1126 conclusion. This must not happen, we want to preserve it under any
1127 circumstances. I will have to write tests with overlong sequences that
1128 get lost in arithmetic and must see if all still works well. DONE
1130 But one fragile point remains: if one host considers a>b and the other
1131 one considers them == but no eq. To prevent this, we must probably do
1132 some extra homework. DONE
1134 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1136 * the concept of tracking DONE needs an object per recentfile that has
1137 something like these methods:
1139 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1141 covered() register() covered()
1143 The unclear thing is how we translate points in time into intervals. We
1144 could pass a reference to the current recent_events array when running
1145 we_have(xxx) and let the DONE object iterate over it such that it only
1146 has to store a list of intervals that can melt into each other. Ah, even
1147 passing the list together with a list of indexes seems feasiable.
1149 Or maybe ask for the inverted list?
1151 Whenever the complete array is covered by the interval we say we are
1152 fully covered and if the recentfile is not expired, we are uptodate.
1154 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1156 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1158 * need a way to "return" the next entry after the end of a list. When
1159 the caller says "before" or "after" we would like to know if he could
1160 cover that interval/threshold or not because this influences the effect
1161 of a newer timestamp of that recentfile. DONE with $opt{info}.
1163 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1165 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1167 loop implies tixu (?). Nope, something like --statefile decides. Per
1168 default we do ...?
1170 T test, I init, X including removals, U nodirtymark
1172 So we have no concept of dirtymarks, we only trust that since we are
1173 running we have observed everything steadily. But people will not let
1174 this program run forever so we must consider both startup penalty and
1175 book keeping for later runs. We keep this for later. For now we write a
1176 long running mirror that merges several intervals.
1178 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1180 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1181 the intervals!
1183 * rersyncrecent, the script: default to one week. The name of the switch
1184 is --after. Other switches? --loop!
1186 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1188 * need a switch --skip-deletes (?)
1190 * need a switch --enduser that tells us that the whole tempfile
1191 discipline is not needed when there is no downstream user. (?)
1193 Without this switch we cannot have a reasonable recent.pl that just
1194 displays the recent additions. Either we accept to download everything.
1195 Or we download temporary files without the typical rsync protocol
1196 advantages.
1198 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1199 File::Temp::tempdir, this might be a win.
1201 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1203 * apropos missing: we have no push, we never know the downstream
1204 servers. People who know their downstream hosts and want to ascertain
1205 something will want additional methods we have never thought about, like
1206 update or delete a certain file.
1208 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1210 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1211 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1212 difference.
1214 * rsync options: it's a bit of a pain that we usually need several rsync
1215 options, like compress, links, times, checksum and that there is no
1216 reasonable default except the original rsync default. I think wee can
1217 safely assume that the rsync options are shared between all recentfile
1218 instances within one recent tree.
1220 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1222 * deletes: if a delete follows an add quickly enough it may happen that
1223 a downstream mirror did not see the add at all! It seems this needs to
1224 be mentioned somewhere. The point here is that even if the downstream is
1225 never missing the principal timeframe it may encounter a "delete" that
1226 has no complimentary "add" anywhere.
1228 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1230 * I suspect the treat of metadata is incorrect during read or something.
1231 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1232 contained more than 6 hours worth of data. At 06:08 we merged into the
1233 1d file. We need to take snapshots of the 6h file over the course of an
1234 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1235 enough.
1237 Much worse: watching the 1h file: right at the moment (at 06:35) it
1238 covers 1218867584-1219120397 which is 70 hours.
1240 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1241 08:19:44 2008, that is when I checked out last time, so it seems to be
1242 aggregating and never truncating?
1244 No, correct is: it is never truncating; but wrong is: it is aggregating.
1245 It does receive a lot of events from time to time from a larger file.
1246 Somehow a large file gets merged into the small one and because the
1247 "meta/merged" attribute is missing, nobody is paying attention. I
1248 believe that I can fix this by making sure that metadata are honoured
1249 during read. DONE and test adjusted.
1251 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1253 * grand renaming plan
1255 remotebase => remoteroot to fit well with localroot DONE
1256 local_path() => localroot seems to me should already work DONE
1257 recentfile_basename => rfilename no need to stress it has no slash DONE
1259 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1260 kern? It would anyway need a deprecation cycle because it is an
1261 important constructor.
1263 * I like the portability that Data::Serializer brings us but the price
1264 is that some day we might find out that it is slowing us a bit. We'll
1265 see.
1267 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1269 * should we not enter the interval of the principal (or the interval of
1270 the merging file?) in every aggregated/merged file?
1272 * we should aim at a first release and give up on thinking about
1273 sanitizing stuff and zloop. Let's just admit that a full traditional
1274 rsync is the only available sanitizer ATM. Otherwise it's complicated
1275 stuff: sanitizing on the origin server, sanitizing on the slaves,
1276 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1277 get the basics out before this becomes a major cause for mess.
1279 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1281 * On OSes not supporting symlinks we expect that RECENT.recent contains
1282 the contents of the principal recentfile. Actually this is identical on
1283 systems supporting symlinks. Simple, what follows from that is that we
1284 need to keep the serializer in the metadata because we cannot read it
1285 from the filename, doesn't it? Of course not. It's a chicken and egg
1286 problem. This leaves us with the problem to actually parse the
1287 serialized data to find out in which format it is. So who can do the 4
1288 or 5 magics we wanted to support? File::LibMagic?
1290 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1292 * remotebase and recentfile_basename are ugly names. Now that we need a
1293 word for the shortest/principal/driving recentfile too we should do
1294 something about it.
1296 localroot is good. rfile is good. local_path() is bad, local_path($path)
1297 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1298 already deprecated.
1300 Up to now remotebase was the string that described the remote root
1301 directory in rsync notation, like pause.perl.org::authors. And
1302 recentfile_basename was "RECENT-1h.yaml".
1304 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1306 * The test that was added in today's checkin is a good start for a test
1307 of rmirror. We should have more methods in Recent.pm: verify,
1308 addmissingfiles. We should verify the current tree, then rmirror it and
1309 then verifytree the copy. We could then add some arbitrary file and let
1310 it be discovered by addmissingfiles, then rmirror again and then
1311 verifytree the copy again.
1313 Then we could start stealing from csync2 sqlite database [no port to
1314 OSX!] and fill a local DB. And methods to compare the database with the
1315 recentfiles. Our strength is that in principle we could maintain state
1316 with a single float. We have synced up to 1234567890.123456. If the Z
1317 file does not add new files all we have to do is mirror the new ones and
1318 delete the goners.
1320 This makes it clear that we should extend current protocol and declare
1321 that we cheat when we add files too late, just to help the other end
1322 keeping track. Ah yes, that's what was meant when zloop was mentioned
1323 earlier.
1325 Maybe need to revisit File::Mirror to help me with this task.
1327 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1329 * There must be an allow-me-to-truncate flag in every recentfile.
1330 Without it one could construct a sequence of updates winning the locking
1331 battle against the aggregator. Only if an aggregator has managed to
1332 merge data over to the next level, truncating can be allowed. DONE with
1333 accessor merged.
1335 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1337 * We should probably guarantee that no duplicates enter the aggregator
1338 array.
1340 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1342 * To get merge operation faster would need a good benchmark test. What
1343 02 spits out isn't reliable enough and is dominated by many other
1344 things. Between
1346 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1347 Author: Andreas J. Koenig <k@k75.(none)>
1348 Date: Sat Aug 2 07:58:04 2008 +0200
1350 and
1352 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1353 Author: Andreas J. Koenig <k@k75.(none)>
1354 Date: Sat Aug 2 11:40:29 2008 +0200
1356 there was a lot of trying but the effect is hardly measurable with
1357 current tests.
1359 * overhead of connecting seems high. When setting
1360 max_files_per_connection to 1 we see that.
1362 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1364 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1366 25138.2215600014/3600 = 6.98283932222261
1368 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1369 be a bug in the aggregator. FIXED
1371 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1373 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1374 we have it, can we mirror it, mirror it!
1376 I fear this needs a new class which might be called
1377 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1378 and treat them as an entity. I realize that a single recentfile may be
1379 sufficient for certain tasks and that it is handy for the low level
1380 programmer but it is not nice to use. If there is a delete in the 1h
1381 file then the 6h file still contains it. Seekers of the best information
1382 need to combine at least some of the recentfiles most of the time.
1384 There is the place for the Z loop!
1386 But the combination is something to collect in a database, isn't it. Did
1387 csync2 just harrumph?
1389 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1391 * it just occurred to me that hosts in the same mirroring pool could
1392 help out each other even without rewriting the recentfile. Just fetch
1393 the stuff to mirror from several places, bingo. But that's something
1394 that should rather live in a separate package or in rsync directly.
1396 * cronjobs are unsuited because with ntp they would all come at the full
1397 minute and disturb each other. Besides that I'd hate to have a backbone
1398 with more than a few seconds latency.
1400 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1402 * a second rsync server with access control for PAUSE. Port? 873 is the
1403 standard port, let's take 8873.
1405 * if there were a filesystem based on this, it would have a slow access
1406 to inexistent files. It would probably provide wrong readdir (only based
1407 on current content) or also a slow one (based on a recentfile written
1408 after the call). But it would provide fast access to existing files. Or
1409 one would deliberately allow slightly blurred answers based on some
1410 sqlite reflection of the recentfiles.
1412 * todo: write a variant of mirror() that combines two or more
1413 recentfiles and treats them like one
1415 * todo: signal handler to remove the tempfile
1417 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1419 * now that we have the symlink I forgot how it should be used in
1420 practice.
1422 * the z loop: add missing files to Z file. Just append them (instead of
1423 prepending). So one guy prepends something from the Y file from time to
1424 time and another guy appends something rather frequently. Collecting
1425 pond. When Y merges into Z, things get epoch and the collecting pond
1426 gets smaller. What exactly are "missing files"?
1428 take note of current epoch of the alpha file, let's call it the
1429 recent-ts
1431 find all files on disk
1433 remove all files registered in the recentworld up to recent-ts
1435 remove all files that have been deleted after recent-ts according to
1436 recentworld
1438 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1440 * rersyncrecent might be a cronjob with a (locked) state file which
1441 contains things like after and maybe last z sync or such?
1443 rrr-mirror might be an alternative name but how would we justify the
1444 three Rs when there is no Re-Rsync-Recent?
1446 With the --loop parameter it is an endless loop, without it is no loop.
1447 At least this is simple.
1449 * todo: new accssor z-interval specifies how often the Z file is updated
1450 against the filesystem. We probably want no epoch stamp on these
1451 entries. And we want to be able to filter the entries (e.g. no
1452 by-modules and by-category tree)
1454 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1456 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1457 into Z. We do not need the other files filled up with filesystem
1458 contents though.
1460 * need interface to query for a file in order to NOT call update on
1461 PAUSE a second time within a short time.
1463 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1465 * recommended update interval? Makes no sense, is different for
1466 different users.
1468 * Moosify
1470 Local Variables:
1471 mode: change-log
1472 change-log-default-name: "Todo"
1473 tab-width: 2
1474 left-margin: 2
1475 End: