| blob | fa2f7869d5060b2698d6bd902316285bc659bd13 |
1 2011-02-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * minor bug: fsck just added the lockfile to the index which should be
4 considered bookkeeping.
6 * speed: a large directory remove that is done by the kernel in 1
7 seconds triggers minutes of bookkeeping (07:56:42 - 08:01:10 for ~3000
8 files, all due to Schlemiehl again. Want to do some collecting of
9 pending ops before actually writing to the RECENT files. Lock!
11 Repeating the timing with 2600 files and it took 21:28:08 - 21:35:27.
13 Now having rewritten the loop to use batch_update(): 4383 file removed
14 21:43:37 - 21:44:12. From 7:30 to 0:35 while doing 60% more work that's
15 a joy.
17 * lockdirectory expiration? server died and blocked fsck for so long.
19 * At the moment rrr-init is not needed anymore, rrr-server can be
20 started and followed by rrr-fsck. Still stupid but not the main
21 showstopper.
23 2011-02-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
25 * IN_Q_OVERFLOW needs to trigger an fsck, likewise the entry into the
26 server.
28 2011-01-30 Andreas J. Koenig <andk@cpan.org>
30 * rrr-init and rrr-server need to be used such:
32 init
33 server
34 init again
36 Stupid but no showstopper ATM.
38 2010-10-26 Andreas J. Koenig <andk@cpan.org>
40 * forking slave is done, needs more testing.
42 2010-07-06 Andreas J. Koenig <andk@cpan.org>
44 * The two next most important goals are to get a useable rrr-server and
45 a forking slave implementation that uses virtually no memory after the
46 first pass through. Does it need the status file?
48 * rrr-init now being fast we can now remove the alpha alerts. Done.
50 2010-06-16 Andreas J. Koenig <andk@cpan.org>
52 * rrr-init seems to be sssslllloooowwww. At 3:49 I started it for the
53 video tree. After twelve minutes the file has reached a size of 500k. It
54 is locking and rewriting the YAML after every file. Wasn't there an
55 unlocked variant? For "init" one would not expect locked correctness and
56 a rewrite after every item. Given that this tends to become slower with
57 every file it is really unacceptable.
59 Reminds me of rrr-dirtyupdate (the former name of rrr-update). It seems
60 the code was not dirty enough and that was the reason why it was so
61 slow. And rrr-update does not have a --dirty switch, so it looks like we
62 have no way to force dirty and fast operation yet but we need it for
63 init.
65 Irritating the bootstrapping race: the order {(1) fill initial
66 array, (2) start server} leaves the need for an fsck to cover the time
67 between 1 and 2. Given that regular fsck is needed in any case the price
68 is not high but still doesn't give a warm feeling. Maybe starting the
69 server will by itself do an fsck?
71 2010-06-15 Andreas J. Koenig <andk@cpan.org>
73 * remove the wild alpha alerts and say, it is alpha but parts of it are
74 already working very well or so.
76 2009-12-10 Andreas J. Koenig <andk@cpan.org>
78 * Prominent request for statusfile to safe memory consumption. I'd like
79 to work backwards: get a verification program that verifies how well in
80 sync two nodes are without using frmr itself. Pretend that the forking
81 agent already works and can be interrupted with ^C. Then let it die at
82 random and restart at random and see the memory consumption and the
83 failure rate over time. Compare it with the curve we have now.
85 Of course we also invented the status file as a sharp debugging tool.
86 I'd like to see it in action. See also: F:R:M:Recent::thaw().
88 2009-06-22 Andreas J. Koenig <andk@cpan.org>
90 * Todo: rrr-init to generate some index files so that rrr-server can be
91 started.
93 2009-05-02 Andreas J. Koenig <andk@cpan.org>
95 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
96 so we can use it for normal update with a single file. And add a cronjob
97 that runs rrr-fsck so we get a reminder when something is not working
98 as it should.
100 2009-04-27 Andreas J. Koenig <andk@cpan.org>
102 * Todo: I still see more link_stat errors than expected. I come back to
103 the former suspicion that we do not do necessary bookkeeping on deleted
104 files so that when we reach the "new" event on a meanwhile deleted file
105 we try to mirror it although we could know better. It's just noise, no
106 harm but noise is irritating.
108 * Todo: more experimenting with runstatusfile to get it official
110 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
111 parse the error.
113 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
115 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
117 And that ignore_link_stat_error is propagated in sparse_clone.
119 DONE
121 For the record: the bug was that ignore_link_stat_error was lost during
122 sparse_clone. And then the second bug was that it did not default to
123 true. Both are now fixed.
125 2009-04-25 Andreas J. Koenig <andk@cpan.org>
127 * Todo: fill rrr program with life
129 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
131 * Minor bug: I see complaints about files not existing like
132 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
133 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
135 In the RECENT files these are delete events but of course also new
136 events in older recent files.
138 I think we do not properly remember deletes when we skip-deletes and so
139 do not filter them out when we later see the "new" event.
141 Yes, but that what ignore_link_stat_error is for, we know about this
142 race condition. We must make ignore_link_stat_error default to true.
144 DONE
146 * bug in Done.pm failing to merge two fields into one when a third is
147 present. FIXED and accompanied with new test.
149 2009-04-24 Andreas J. Koenig <andk@cpan.org>
151 * Todo: keep the index clean and run some kind of nondestructive fsck 4
152 times a day.
154 * who is our backbone?
156 http://cpan.cpantesters.org/authors/02STAMP # barbie
157 http://cpan.solfo.com/authors/02STAMP # abh
158 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
160 2009-04-17 Andreas J. Koenig <andk@cpan.org>
162 * the changelog helper I included in the Makefile release memo does not
163 show tags and is much less useful than I thought. gitk is probably much
164 more convenient.
166 2009-04-16 Andreas J. Koenig <andk@cpan.org>
168 * ABH writes:
170 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
171 #### Ydr_.yaml ... DONE
172 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
173 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
174 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
175 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
176 #### gone already at cpan-pause.pl line 0
178 Running without skip-deletes now but cannot reproduce.
180 2009-04-15 Andreas J. Koenig <andk@cpan.org>
182 * consider state directory such we can restart after a ^C or come again
183 from a cronjob.
185 Reuse _runstatusfile: write atomically, see how we can use the dump to
186 start running again. Do not leak and remove globs before writing. Do not
187 only dump one rf, dump the whole complex.
189 * consider the effect of ^C. Is it safe?
191 * want to use Perl::Version to up the versions on every release. Only if
192 things have changed, of course. Does this actually work?
194 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
195 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
197 Must do the bookkeeping myself when this should happen, maybe based on git?
199 I think I prefer the same version number for all pm files. But one also
200 needs a routine for setting them. Ahh:
202 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
204 NEEDSMOREWORK
206 * why indexer wrong?
208 cpan[2]> m /Mirror::Recent/
209 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
210 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
211 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
212 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
213 4 items found
215 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
217 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
219 I just tried MM 6.5102 and it again produces only a provides for Recent,
220 not for the other modules.
222 My own fault. FIXED in the Makefile.PL.
224 2009-04-13 Andreas J. Koenig <andk@cpan.org>
226 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
228 * is it true (as stated at
229 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
230 that rsync can lead to errors when upstream changes before client sync
231 has completed? Can we see the error when we run this on upstream:
233 % perl -e '
234 use Time::HiRes qw(time);
235 while (){
236 open my $fh, ">", "changingfile.txt.new" or die;
237 print $fh (time."\n") x 1000000;
238 close $fh;
239 rename "changingfile.txt.new", "changingfile.txt" or die;
240 }
241 '
243 And on the receiving end:
245 % while true; do
246 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
247 sleep 1;
248 done
250 I cannot get it to fail. Apparently it is sufficient when upstream
251 always writes atomically (which of course is mandatory)?
253 SENT to mailing list Apr 28.
255 * from ABH:
257 > https://fedorahosted.org/InstantMirror/
258 > https://www.redhat.com/mailman/listinfo/instantmirror-list
259 > irc.freenode.net channel #instantmirror
261 A cool name for a project. Inspires me to write a few sentences about
262 bittorrent's role in the grand picture.
264 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
265 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
267 2009-04-12 Andreas J. Koenig <andk@cpan.org>
269 * Interesting last minute bug during real download testing: the output
270 isn't as pretty anymore, it seems that more work is being done than
271 needed.
273 Introducing a highlevel _alluptodate method helped a lot in debugging
274 this.
276 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
277 different on 1h file than on the other files. This lead to frequent
278 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
279 issue.
281 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
283 * Todo: try if we can be faster with a native float package. I'm really
284 glad to have a machine exposing perl floating point bugs, it forced me
285 to retract this item from the changes file: "several speedups in the
286 fakebigfloat code" before I get cpan testers fails.
288 2009-04-11 Andreas J. Koenig <andk@cpan.org>
290 * Bug: when the dirtymark changes upstream, then the downstream server
291 notices it immediately and mirrors the "1h" file but then leaves the
292 rmirror loop and continues with "6h" on the next round through. This
293 behaviour goes on until it reaches the Z file. Because we eat the Z file
294 piecemeal we leave the loop after a while but after that everybody seems
295 to have forgotten that there is some work left to be done. Bad, bad,
296 bad.
298 ETOOCONFUSING better name?
300 Recentfile::get_remote_recentfile_as_tempfile OK
301 Recentfile::resolve_recentfilename split_rfilename DONE
303 Recent::_principal_recentfile_object _principal_recentfile_fromremote
304 Recent::principal_recentfile OK
305 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
307 some void, some not, some not void but called in void context.
309 2009-04-10 Andreas J. Koenig <andk@cpan.org>
311 * Bug? When the dirtymark changes then the second tier hosts quickly
312 reset their DONE state and restart mirroring. But then they mirror
313 potentially outdated index files with inconsistent dirtymark because the
314 first tier box may be a bit behind on the large recentfiles.
316 This means we need a brake when iterating over the recentfiles that
317 refuses to use a recentfile with the wrong dirtymark.
319 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
321 * install the bin/* files and talk about rrr-overview somewhere.
323 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
324 created outside the target tree.
326 * Barbie asks for better logging capabilities
328 2009-03-29 Andreas J. Koenig <andk@cpan.org>
330 * known bug: cannot configure the Recentfile objects to keep delete
331 events. How to call the accessor for that? keep_delete_objects_forever?
333 Sounds acceptable. FIXED (but untested).
335 * known bug: keeps temporary index files lying around.
337 2009-03-24 Andreas J. Koenig <andk@cpan.org>
339 * Some equivalent for
341 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
343 ? DONE
345 2009-03-22 Andreas J. Koenig <andk@cpan.org>
347 * k81 is again client of k75 with different parameters.
349 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
350 about it all the time and nobody reacts accordingly; (2) lots of temp
351 files get created and not removed; culprit the new call to
352 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
353 caller has to remove the tempfile after use.
355 Need the drawing board.
357 PARTLY FIXED: retracting the idea to call
358 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
359 and unseed such that uptodate now means "we have mirrored this
360 recentfile's bunch of files" and seeded means "somebody believes the
361 index file for this rf needs be refreshed". This seems to work now.
363 Still collecting temp files that nobody cares to depollute.
365 2009-03-21 Andreas J. Koenig <andk@cpan.org>
367 * Bug?: should it be harder than it is atm to set the timestamp to the
368 future? YES, FIXED
370 * bug with native integers:
372 -
373 epoch: 997872011
374 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
375 type: new
376 -
377 epoch: 1195248431
378 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
379 type: new
381 Native integer broke when native math was turned off. FIXED
383 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
384 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
385 in this area.
387 Yes, records out of order:
389 447003 -
390 447004 epoch: 995885533
391 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
392 447006 type: new
393 447007 -
394 447008 epoch: 995890358
395 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
396 447010 type: new
397 447011 -
398 447012 epoch: 995892221
399 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
400 447014 type: new
402 FIXED with sanity check and later with the integer fix.
404 * Bug: want the index files in a .recent directory
406 * Bug: lots of dot files are not deleted in time
408 * possible test case: can a delete change the timestamp? This would
409 probably break the order of events.
411 2009-03-20 Andreas J. Koenig <andk@cpan.org>
413 * 1233701831.34486 what's so special about this number/string? It
414 belongs to
415 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
416 atm lives in Y,Q, and Z.
418 It is the first entry after id/--skip-locking which has timestamp
419 1234164228.11325 which represents a file that doesn't exist anymore.
421 * another bug:
423 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
424 _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
425 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
426 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
427 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
428 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
429 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
430 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
431 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
432 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
433 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
434 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
435 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
437 FIXED, it was the "--skip-locking" file where manual intervention was participating
439 * bug on the mirroring slave: when the dirtymark gets increased we
440 probably do not reset the done intervals. The mirrorer stays within
441 tight bounds where it tries to sync with upstream and never seems to
442 finish. In the debugging state file I see lots of identical intervals
443 that do not get collapsed. When I restart the mirrorer it dies with:
445 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
446 _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
447 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
448 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
449 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
450 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
451 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
452 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
453 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
454 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
456 and debugging stands at
458 At: before
459 Brfinterval: Z
460 Ci: 227
461 Dre+1:
462 epoch: 1237400802.5789
463 path: id/J/JO/JOHND/CHECKSUMS
464 type: new
465 Dre-0:
466 epoch: 1237400807.97514
467 path: id/M/MI/MIYAGAWA/CHECKSUMS
468 type: new
469 Dre-1:
470 epoch: 1237400817.94363
471 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
472 type: new
473 Eintervals:
474 -
475 - 900644040
476 - 900644040
477 - []
479 and it is reproducable.
481 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
482 pause has a fresh one.
484 FIXED, it was the third anded term in each of the ifs in the IV block in
485 _register_one: with that we make sure that we do not stamp on valuable
486 interval data.
488 2009-03-17 Andreas J. Koenig <andk@cpan.org>
490 * done: verified the existence of the floating point bug in bleadperl
491 and verified that switching from YAML::Syck to YAML::XS does not resolve
492 it.
494 BTW, the switch was doable with
496 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
498 and should be considered as a separate TODO
500 * todo: integrate a dirty update with two aggregate calls before
501 unlocking for frictionless dirtying DONE
503 * todo: start the second rsync daemon on pause DONE
505 * todo: move index files to .recent: this cannot simply be done by
506 setting filenameroot to .recent/RECENT. Other parts of the modules rely
507 on the fact that dirname(recentfile) is the root of the mirrored tree.
509 2009-03-16 Andreas J. Koenig <andk@cpan.org>
511 * What was the resolution of the mirror.pl delete hook bug? Do we call
512 the delete hook when pause removes a file from MUIR?
514 * Today on pause: Updating 2a13fba..29f284d and installing it for
515 /usr/local/perl-5.10.0{,-RC2}
517 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
518 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
520 2009-03-15 Andreas J. Koenig <andk@cpan.org>
522 * currently recent_events has the side effect of setting dirtymark
523 because it forces a read on the file. That should be transparent, so
524 that the dirtymark call always forces a cache-able(?) read.
526 * The bug below is -- after a lot of trying -- not reproducible on a
527 small script, only in the large test script. The closest to the output
528 below was:
530 #!perl
531 use strict;
532 use Devel::Peek;
533 my $x = "01237123229.8814";
534 my($l,$r);
535 for ($l,$r) {
536 $_ = "x"x34;
537 }
538 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
539 $r = int $r;
540 $l = "1237123231.22458";
541 $r = "1237123231.22458";
542 1 if $l/1.1;
543 Devel::Peek::Dump $l;
544 Devel::Peek::Dump $r;
545 Devel::Peek::Dump $x = $l <=> $r;
546 die "BROKE" if $x;
547 __END__
549 The checked in state at c404a85 fails the test with my
550 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
551 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
552 in the test script always consistent.
554 * Todo: write a test that inserts a second dirty file with an already
555 existing timestamp. DONE
557 * Bug in perl 5.10 on my 64bit box:
559 DB<98> Devel::Peek::Dump $l
560 SV = PVMG(0x19e0450) at 0x142a550
561 REFCNT = 2
562 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
563 IV = 0
564 NV = 1237123231.22458
565 PV = 0x194ce70 "1237123231.22458"\0
566 CUR = 16
567 LEN = 40
569 DB<99> Devel::Peek::Dump $r
570 SV = PVMG(0x19e0240) at 0x142a3e8
571 REFCNT = 2
572 FLAGS = (PADMY,POK,pPOK)
573 IV = 1237123229
574 NV = 1237123229.8814
575 PV = 0x19ff900 "1237123231.22458"\0
576 CUR = 16
577 LEN = 40
579 DB<100> Devel::Peek::Dump $l <=> $r
580 SV = IV(0x19ea6e8) at 0x19ea6f0
581 REFCNT = 1
582 FLAGS = (PADTMP,IOK,pIOK)
583 IV = -1
585 DB<101> Devel::Peek::Dump $l
586 SV = PVMG(0x19e0450) at 0x142a550
587 REFCNT = 2
588 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
589 IV = 0
590 NV = 1237123231.22458
591 PV = 0x194ce70 "1237123231.22458"\0
592 CUR = 16
593 LEN = 40
595 DB<102> Devel::Peek::Dump $r
596 SV = PVMG(0x19e0240) at 0x142a3e8
597 REFCNT = 2
598 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
599 IV = 1237123231
600 NV = 1237123231.22458
601 PV = 0x19ff900 "1237123231.22458"\0
602 CUR = 16
603 LEN = 40
605 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
607 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
609 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
611 * $obj->merge ($other) needs to learn about equal epoch which may happen
612 since dirty_epoch intruded.
614 * Wontfix anytime soon: I think we currently do not support mkdir. Only
615 files!
617 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
619 * Todo: continue working on update(...,$dirty_epoch). It must be
620 followed by a fast_aggregate! DONE
622 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
624 * maybe we need a closest_entry or fitting_interval or something like
625 that. We want to merge an event into the middle of some recentfile.
626 First we do not know which file, then we do not know where to lock,
627 where to enter the new item, when and where to correct the dirtymark.
629 So my thought is we should first find which file.
631 Another part of my brain answers: what would happen if we would enter
632 the new file into the smallest file just like an ordinary new event,
633 just as an old event?
635 (1) we would write a duplicate timestamp? No, this would be easy to
636 avoid
638 (2) we would make the file large quickly? Yes, but so what? We are
639 changing the dirtymark, so are willing to disturb the downstream hosts.
641 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
643 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
645 leak!
647 https://rt.cpan.org/Ticket/Display.html?id=41199
649 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
651 Strangest output being HAKANARDO who managed to upload
653 Here is a better oneliner that includes also the first line of each
654 finding:
656 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
658 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
660 * memory leak in the syncher? It currently weighs 100M.
662 Update 2008-11-02:
664 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
667 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
669 * lookup by epoch and by path and use this ability on the pause to never
670 again register a file twice that doesn't need it. Let's call it
671 contains().
673 * after the dirtymark is done: fill up recentfiles with fake (historic)
674 entries; fill up with individual corrections; algorithm maybe to be done
675 with bigfloat so that we can always place something in the middle
676 between two entries. Before we must switch to bigfloat we could try to
677 use Data::Float::nextup to get the.
679 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
680 directly.
682 * dirtymark now lives in Recentfile, needs to be used in rmirror.
684 * find out why the downloader died after a couple of hours without a net
685 connection. Write a test that survives the not-existence of the other
686 end forever.
688 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
690 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
691 it makes less sense with tight coupling for secondary files. But asking
692 the server what the current epoch is might be cheaper on HTTP than on
693 rsync. (Needs to be evaluated)
695 * remove the 0.00 from the verbose overview in the Merged column in the
696 Z row. DONE
698 * write tests that expose the problems of the last few days: cascading
699 client/server roles, tight coupling for secondary RFs, deletes after
700 copies.
702 * Some day we might want to have policy options for the slave:
703 tight/loose/no coupling with upstream for secondary RFs. tight is what
704 we have now. loose would wait until a gap occurs that can be closed.
706 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
708 * revisit all $rfs->[$i+1] places if they now make sense still
710 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
712 * another bug is the fact that the mirror command deletes files before
713 it unhides the index file, thus confusing downstream slaves. We must not
714 delete before unhiding and must delete after unhiding. FIXED.
716 * new complication about the slave that is playing a server role.
717 Currently we mirror from newest to oldest with a hidden temporary file
718 as index. And when one file is finished, we unhide the index file.
719 Imagine the cascading server/slave is dead for a day. It then starts
720 mirroring again with the freshest thing and unhides the freshest index
721 file when it has worked through it. In that moment it exposes a time
722 hole. Because it now works on the second recentfile which is still
723 hidden.
725 We currently do nothing special to converge after such a drop out. At
726 least not intentionally and robustly and thought through.
728 The algorithm we use to seed the next file needs quite a lot of more
729 robustness than it currently has. Something to do with looking at the
730 merged element of the next rf and when it has dropped off, we seed
731 immediately. And if it ramains dropped off, we seed again, of course.
733 Nope, looking from smaller to larger RFS we look at the merged element
734 of this RF and at the minmax/max element of the next RF. If that
735 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
736 seeding it.
738 And we need a public accessor seed and unseed or seeded. But not the mix
739 of public and private stuff that then is used behind the back.
741 And then the secondary* stuff must go.
743 And we must understand what the impact is on the DONE system. Can it go
744 unnoticed that there was a hole? And could the DONE system have decided
745 the hole is covered? This should be testable with three directories where
746 the middle stops working for a while. Done->merge is suspicious, we must
747 stop it from merging non-conflatable neighbors due to broken continuity.
749 FIXED
751 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
753 * Slaven suggests to have the current epoch or the whole current
754 recentfile available from the HTTP server and take it away with
755 keepalive. This direction goes the granularity down to subseconds.
757 We might want to rewrite everything to factor out transport and allow
758 the whole thing to run via HTTP.
760 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
762 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
763 symlink?
765 * smoker on k81 fetching from k75 to verify cascading works. See
766 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
768 * maybe the loop should wait for CHECKSUMS file after every upload. And
769 CPAN.pm needs to deal with timestamps in the future.
771 * do not forget the dirtymark!
773 Text: have a new flag on recentfiles with the meaning: if this
774 changes, you're required to run a full rsync over all the files. The
775 reason why we set it would probably be: some foul happened. we injected
776 files in arbitrary places or didn't inject them although they changed.
777 The content of the flag? Timestamp? The relation between the
778 recentfiles would have to be inheritance from the principal, because any
779 out of band changes would soon later propagate to the next recentfile.
781 By upping the flag often one can easily ruin the slaves.
783 last out of band change? dirtymark?
785 Anyway, this implies that we read a potentially existing recentfile
786 before we write one.
788 And it implies that we have an eventloop that keeps us busy in 2-3
789 cycles, one for current stuff (tight loop) and one for the recentfiles
790 (cascade when principal has changed), one for the old stuff after a
791 dirtymark change.
793 And it implies that the out-of-band change in any of the recentfiles
794 must have a lock on the principal file and there is the place to set the
795 dirtymark.
797 * start a FAQ, especially quick start guide questions. Also to aid those
798 problematic areas where we have no good solution, like the "links"
799 option to rsync.
801 * wish feedback when we are slow.
803 * reduce mccabe
805 * Remove a few DEBUG statements.
807 * The multiple-rrr way of doing things needs a new option to rmirror,
808 like piecemeal or so. Not urgent because after the first pass through,
809 things run smoothely. It's only ugly during the first pass.
811 * I have the suspicion that the code is broken that decides if the
812 neighboring RF needs to be seeded. I fear when too much time has gone
813 between two calls (in our case more than one hour), it would not seed
814 the neighbor. Of course this will never be noticed, so we need a good
815 test for it.
817 * local/localroot confusion: I currently pass both options but one must
818 do.
820 * accounts for early birds on PAUSE rsync daemon.
822 * hardcoded 20 seconds
824 * who mirrors the index? DOING now.
826 * which CPAN mirrors offer rsync?
828 * visit all XXX, visit all _float places
830 * rename the pathdb stuff, it's too confusing. No idea how.
832 * rrr-inotify, backpan, rrr-register
834 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
836 * current bugs: the pathdb seems to get no reset, the seeding of the
837 secondaryttl stuff seems not to have an effect. Have helped myself with
838 a rand(10), need to fix this back. So not checked in. Does the rand
839 thing even help?
841 The rand thing helps. The secondaryttl stuff was in the wrong line,
842 fixed now.
844 The pathdb stuff was because I called either _pathdb or __pathdb on the
845 wrong object. FIXED now.
847 * It's not so beautiful if we never fetch the recentfiles that are not
848 the principal, even if this is correct behaviour. We really do not need
849 them after we have fetched the whole content.
851 OK, we want a switch for that: secondaryttl DONE
853 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
855 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
856 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
857 it lets 40 follow. FIXED
859 * See that the long running process really only updates the principal
860 file unless it has missed a timespan during which something happened. If
861 nothing happened, it must notice even when it misses the timespan. DONE
863 * we must throw away the pathdb when we have reached the end of Z. From
864 that moment we can have a very small pathdb because the only reason for
865 a pathdb is that we know to ignore old records in old files. We won't
866 need this pathdb again before the next full pass over the data is
867 necessary and then we will rebuild it as we go along. DONE
869 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
871 * I think, Done::register_one is doing wrong in that it does not
872 conflate neighboring pieces. The covered() method cannot do this because
873 it has no recent_events array at hand. But register_one has it and could
874 do it and for some reason misses to do it (sometimes).
876 This means that the three tests I just wrote can probably not survive
877 because they test with an already broken Done structure.
879 The art now is to detect how it happens, then to reproduce, then write a
880 test, then fix it.
882 So from the logfile this is what happens: we have a good interval with
883 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
884 top of it. Locally we now mirror F2 and open a new done interval for it.
885 Then we mirror F1 but this time with the timestamp T1b. And when we then
886 try to close the gap, we do not find T1 but instead something older. We
887 should gladly accept this older piece and this would fix this bug.
889 FIXED
891 * bug to fix: when the 1h file changes while rmirror is running, we do
892 correctly sync the new files but never switch to the 6h file but rather
893 stay in a rather quick loop that fetches the 1h file again and again.
895 Is it possible that we initialize a new object? Or does
896 get_remote_recentfile_as_tempfile overwrite something in myself?
898 Want a new option: _runstatusfile => $file which frequently dumps the
899 state of all recentfiles to a file.
901 FIXED
903 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
905 * Todo: now teach update to verify the timestamp is about to write
906 against the previous and use _increase_a_bit if it doesn't comply with
907 strict monotony. DONE
909 * The problem of rounding. So far perl's default precision was
910 sufficient. One day it won't be. FakeFloat has an easy job when it is
911 only reading and other machines have written correctly. But when we want
912 to write a floating point number that is a bit larger than the other
913 one, then we need our own idea of precision.
915 Slaven said: just append a "1". This might be going towards the end of
916 usability too quickly. I'd like something that actually uses the decimal
917 system. Well, appending a 1 also does this but...
919 E.g. we have 1.0. nextup on this architecture is starting with
920 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
921 taken the 1.0000000000000003 and the next user comes and the time tells
922 him 1.0 again. He has to beat my number without stepping over the
923 nextup. This is much less space than I had when I chose 1,2,3.
925 What is also irritating is that nextup is architecture dependent. The
926 128 bit guy must choose very long numbers to fit in between whereas the
927 other one with 16 bit uses larger steps. But then the algorithm is the
928 same for both, so that would be a nice thing.
930 I see two situation where we need this. One is when Time::HiRes returns
931 us a value that is <= the last entry in our recentfile. In this case
932 (let's call it the end-case) we must fill the region between that number
933 and the next higher native floating point number. The other is when we
934 inject an old file into an old recentfile (we would then also set a new
935 dirtymark). We find the integer value already taken and need a slightly
936 different one (let's call it the middle-case). The difference between
937 the two situations is that the next user will want to find something
938 higher than my number in the end-case and something lower than my number
939 in the middle case.
941 So I suggest we give the function both a value and an upper bound and it
942 calculates us a primitive middle. The upper bound in the middle-case is
943 the next integer. The upper bound on the end-case is the nextup floating
944 point number. But the latter poses another problem: if we have occupied
945 the middle m between x and nextup(x), then the nextup(m) will probably
946 not be the same as nextup(x) because some rounding will take place
947 before the nextup is calculated and when the rounding reaches the
948 nextup(x), we will end up at nextup(nextup(x)).
950 So we really need to consider the nextup and the nextdown from there and
951 then the middle and that's the number we may approach asymptotically.
952 Ugly. But DONE.
954 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
956 * consider deprecating the use of RECENT.recent as a symlink. It turns
957 out to need extra hoops with the rsync options and just isn't worth it.
958 Or maybe these extra hoops are needed anyway for the rest of the tree?
959 Nope, can't be the case because not all filesystems support symlinks.
961 But before doing the large step, I'll deprecate the call of
962 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
963 only introduced to resolve RECENT.recent and complicates the routine far
964 beyond what it deserves.
966 DONE. Won't deprecate RECENT.recent, just moved its handling to the
967 supervisor.
969 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
971 * I think it's a bug that the rsync_option links must be set to true in
972 order to support RECENT.recent and that nobody cares to set it
973 automatically. Similar for ignore_link_stat_errors.
975 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
977 * Todo: find all todos together and make a plan what is missing for a
978 release.
980 - verifytree or something like that. fsck maybe.
982 - rersyncrecent, the script itself? What it do?
984 - a way to only mirror the recentfiles without mirroring the whole
985 remote system such that people can decide to mirror only partially see
986 also 2008-08-30. .shadow-xxx directory? this also needed for a
987 filesystem that is still incomplete and might need the mirrorfiles for
988 lookup(?)
990 - long living objects that mirror again and again. Inject something
991 into ta, see how it goes over to tb.
993 - how do we continue filling up the DONE system when we use an object
994 for the second time? "fully covered" and "uptodate" or new terminology.
996 - overview called on the wrong file should be understandable
998 - the meta data field that must change when we fake something up so that
999 the downstream people know they have to re-fetch everything.
1001 - how tolerant are we against missing files upstream? how do we keep
1002 track? there are legitimate cases where we did read upstream index right
1003 before a file got deleted there and then find that file as new and want
1004 it. There are other cases that are not self healing and must be tracked
1005 and bugreported.
1007 - how, exactly, do we have to deal with deletes? With rsync errors?
1009 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
1010 authors) failed: No such file or directory (2)
1012 The file above is a delete in 1h and a new in file 1M and the
1013 delete in the locally running rmirror did not get propagated to the 1M
1014 object. Bug. And the consequence is a standstill.
1016 It seems that a slave that works with a file below the principal needs
1017 to merge things all the way up to get rid of later deletes. Or keep
1018 track of all deletes and skip them later. So we need a trackdeletes.pm
1019 similar to the done.pm?
1021 see also 2008-08-20 about spurious deletes that really have no add
1022 counterpart and yet they are not wrong.
1024 - consider the effect when resyncing the recentfile takes longer than
1025 the time per loop. Then we never rsync any file. We need to diagnose
1026 that and force an increase of that loop time. But when we later are fast
1027 enough again because the net has recovered, then we need to switch back
1028 to original parameters. ERm, no, it's enough to keep syncing at least
1029 one file before refetching an index file.
1031 - remember to verify that no temp files are left lying around and the
1032 signal handler
1034 - status file for not long running jobs that want to track upstream with
1035 a, say, cronjob.
1037 - revisit all XXX _float areas and study Sub::Exporter DONE
1039 - persistent DB even though we just said we do not need it. Just for
1040 extended capabilities and time savings when, for example, upstream
1041 announces a reset and we get new recentfiles and could then limit
1042 ourselves to a subset of files (those that have a changed epoch) in a
1043 first pass and would only then do the loop to verify the rest. Or
1044 something.
1046 * Todo: aggregate files should know their feed and finding the principal
1047 should be done stepwise. (?)
1049 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
1050 left locks around. DONE
1052 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1054 * maybe extend the _overview so that it always says if and where the
1055 last file is in the next file and where the next event in the next rf
1056 would lie. No, don't like this anymore. REJECT
1058 * take the two new redundant tests out again, only the third must
1059 survive. DONE
1061 * Todo: add a sanity check if the merged structure is really pointing to
1062 a different rf and that this different rf is larger. DONE
1064 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1066 * now test, if they are overlapping. And test if there is a file in the
1067 next rf that would fit into this rf's interval.
1069 1h 1222324012.8474 1222322541.7963 0.4086
1070 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
1071 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
1072 1W 1222313218.3626 1221708477.5829 167.9835
1074 I suspect that somebody writes a merged timestamp without having merged
1075 and then somebody else relies on it.
1077 If aggregate is running, the intervals must not be extravagated, if it
1078 is not running, there must not be bounds, the total number of events in
1079 the system must be counted and must be controlled throughout the tests.
1080 That the test required the additional update was probably nonsense,
1081 because aggregate can cut pieces too. FIXED & DONE
1083 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1085 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1086 changed. FIXED
1088 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1090 * Most apparent bug at the moment is that the recentfiles are fetched
1091 too often. Only the principal should be fetched and if it has not
1092 changed, the others should not be refetched. ATM I must admit that I'm
1093 happy that we refetch more often than needed because I can more easily
1094 fix bugs while the thing is running.
1096 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1097 done but the done system does not know about it. The reason for the
1098 failure is not known and we never reach the status uptodate because of
1099 this. We must get over it.
1101 Later it turns out that the origin server had a bug somewhere.
1102 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1103 array of the recentfile one position above. The bug was my own.
1105 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1107 * There is the race condition where the server does a delete and the
1108 slave does not yet know and then tries to download it because he sees
1109 the new. So for this time window we must be more tolerant against
1110 failure. If we cannot download a file, we should just skip it and should
1111 not retry immediately. The whole system should discover the lost thing
1112 later. Keeping track with the DONE system should really be a no brainer.
1114 But there is something more: the whole filesystem is a database and the
1115 recentfiles are one possible representation of it. It's a pretty useful
1116 representation I think that's why I have implemented something around
1117 it. But for strictly local operation it has little value. For local
1118 operation we would much rather have a database. So we would enter every
1119 recentfile reading and every rsync operation and for every file the last
1120 state change and what it leads to. Then we would always ignore older
1121 records without the efforts involved with recentfiles.
1123 The database would have: path,recentepoch,rsyncedon,deletedon
1125 Oh well, not yet clear where this leads to.
1127 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1129 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1130 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1132 * statusfile, maybe only on demand, alone to have a sharp debugging
1133 tool. It is locked and all recentfiles dump themselves into it and we
1134 can build a viewer that lets us know where we stand and what's inside.
1136 * remember: only the principal recentfile needs expiration, all others
1137 shall be expired by principal if it discovers that something has move
1138 upstream.
1140 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1142 * Always check if we stringify to a higher value than in the entry
1143 before. DONE
1145 * And in covered make an additional check if we would be able to see a
1146 numerical difference between the two numbers and if we can't then switch
1147 to a different, more expensive algorithm. Do not want to be caught by
1148 floating surprises. DONE
1150 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1152 * caching has several aspects here: we can cache the interval of the
1153 recentfile which only will change when the mtime of the file changes. We
1154 must re-mirror the recentfile when its ttl has expired. Does have_read
1155 tell you anything? It counts nothing at all. Only the mtime is
1156 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1157 thing the have_mirrored because it is unrelated to the mtime.
1159 * Robustness of floating point calculations! I always thought that the
1160 string calculated by the origin server for the floating representation
1161 of the epoch time is just a string. When we convert it to a number and
1162 later back to a string, the other computer might come to a different
1163 conclusion. This must not happen, we want to preserve it under any
1164 circumstances. I will have to write tests with overlong sequences that
1165 get lost in arithmetic and must see if all still works well. DONE
1167 But one fragile point remains: if one host considers a>b and the other
1168 one considers them == but no eq. To prevent this, we must probably do
1169 some extra homework. DONE
1171 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1173 * the concept of tracking DONE needs an object per recentfile that has
1174 something like these methods:
1176 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1178 covered() register() covered()
1180 The unclear thing is how we translate points in time into intervals. We
1181 could pass a reference to the current recent_events array when running
1182 we_have(xxx) and let the DONE object iterate over it such that it only
1183 has to store a list of intervals that can melt into each other. Ah, even
1184 passing the list together with a list of indexes seems feasiable.
1186 Or maybe ask for the inverted list?
1188 Whenever the complete array is covered by the interval we say we are
1189 fully covered and if the recentfile is not expired, we are uptodate.
1191 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1193 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1195 * need a way to "return" the next entry after the end of a list. When
1196 the caller says "before" or "after" we would like to know if he could
1197 cover that interval/threshold or not because this influences the effect
1198 of a newer timestamp of that recentfile. DONE with $opt{info}.
1200 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1202 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1204 loop implies tixu (?). Nope, something like --statefile decides. Per
1205 default we do ...?
1207 T test, I init, X including removals, U nodirtymark
1209 So we have no concept of dirtymarks, we only trust that since we are
1210 running we have observed everything steadily. But people will not let
1211 this program run forever so we must consider both startup penalty and
1212 book keeping for later runs. We keep this for later. For now we write a
1213 long running mirror that merges several intervals.
1215 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1217 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1218 the intervals!
1220 * rersyncrecent, the script: default to one week. The name of the switch
1221 is --after. Other switches? --loop!
1223 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1225 * need a switch --skip-deletes (?)
1227 * need a switch --enduser that tells us that the whole tempfile
1228 discipline is not needed when there is no downstream user. (?)
1230 Without this switch we cannot have a reasonable recent.pl that just
1231 displays the recent additions. Either we accept to download everything.
1232 Or we download temporary files without the typical rsync protocol
1233 advantages.
1235 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1236 File::Temp::tempdir, this might be a win.
1238 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1240 * apropos missing: we have no push, we never know the downstream
1241 servers. People who know their downstream hosts and want to ascertain
1242 something will want additional methods we have never thought about, like
1243 update or delete a certain file.
1245 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1247 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1248 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1249 difference.
1251 * rsync options: it's a bit of a pain that we usually need several rsync
1252 options, like compress, links, times, checksum and that there is no
1253 reasonable default except the original rsync default. I think wee can
1254 safely assume that the rsync options are shared between all recentfile
1255 instances within one recent tree.
1257 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1259 * deletes: if a delete follows an add quickly enough it may happen that
1260 a downstream mirror did not see the add at all! It seems this needs to
1261 be mentioned somewhere. The point here is that even if the downstream is
1262 never missing the principal timeframe it may encounter a "delete" that
1263 has no complimentary "add" anywhere.
1265 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1267 * I suspect the treat of metadata is incorrect during read or something.
1268 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1269 contained more than 6 hours worth of data. At 06:08 we merged into the
1270 1d file. We need to take snapshots of the 6h file over the course of an
1271 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1272 enough.
1274 Much worse: watching the 1h file: right at the moment (at 06:35) it
1275 covers 1218867584-1219120397 which is 70 hours.
1277 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1278 08:19:44 2008, that is when I checked out last time, so it seems to be
1279 aggregating and never truncating?
1281 No, correct is: it is never truncating; but wrong is: it is aggregating.
1282 It does receive a lot of events from time to time from a larger file.
1283 Somehow a large file gets merged into the small one and because the
1284 "meta/merged" attribute is missing, nobody is paying attention. I
1285 believe that I can fix this by making sure that metadata are honoured
1286 during read. DONE and test adjusted.
1288 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1290 * grand renaming plan
1292 remotebase => remoteroot to fit well with localroot DONE
1293 local_path() => localroot seems to me should already work DONE
1294 recentfile_basename => rfilename no need to stress it has no slash DONE
1296 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1297 kern? It would anyway need a deprecation cycle because it is an
1298 important constructor.
1300 * I like the portability that Data::Serializer brings us but the price
1301 is that some day we might find out that it is slowing us a bit. We'll
1302 see.
1304 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1306 * should we not enter the interval of the principal (or the interval of
1307 the merging file?) in every aggregated/merged file?
1309 * we should aim at a first release and give up on thinking about
1310 sanitizing stuff and zloop. Let's just admit that a full traditional
1311 rsync is the only available sanitizer ATM. Otherwise it's complicated
1312 stuff: sanitizing on the origin server, sanitizing on the slaves,
1313 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1314 get the basics out before this becomes a major cause for mess.
1316 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1318 * On OSes not supporting symlinks we expect that RECENT.recent contains
1319 the contents of the principal recentfile. Actually this is identical on
1320 systems supporting symlinks. Simple, what follows from that is that we
1321 need to keep the serializer in the metadata because we cannot read it
1322 from the filename, doesn't it? Of course not. It's a chicken and egg
1323 problem. This leaves us with the problem to actually parse the
1324 serialized data to find out in which format it is. So who can do the 4
1325 or 5 magics we wanted to support? File::LibMagic?
1327 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1329 * remotebase and recentfile_basename are ugly names. Now that we need a
1330 word for the shortest/principal/driving recentfile too we should do
1331 something about it.
1333 localroot is good. rfile is good. local_path() is bad, local_path($path)
1334 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1335 already deprecated.
1337 Up to now remotebase was the string that described the remote root
1338 directory in rsync notation, like pause.perl.org::authors. And
1339 recentfile_basename was "RECENT-1h.yaml".
1341 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1343 * The test that was added in today's checkin is a good start for a test
1344 of rmirror. We should have more methods in Recent.pm: verify,
1345 addmissingfiles. We should verify the current tree, then rmirror it and
1346 then verifytree the copy. We could then add some arbitrary file and let
1347 it be discovered by addmissingfiles, then rmirror again and then
1348 verifytree the copy again.
1350 Then we could start stealing from csync2 sqlite database [no port to
1351 OSX!] and fill a local DB. And methods to compare the database with the
1352 recentfiles. Our strength is that in principle we could maintain state
1353 with a single float. We have synced up to 1234567890.123456. If the Z
1354 file does not add new files all we have to do is mirror the new ones and
1355 delete the goners.
1357 This makes it clear that we should extend current protocol and declare
1358 that we cheat when we add files too late, just to help the other end
1359 keeping track. Ah yes, that's what was meant when zloop was mentioned
1360 earlier.
1362 Maybe need to revisit File::Mirror to help me with this task.
1364 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1366 * There must be an allow-me-to-truncate flag in every recentfile.
1367 Without it one could construct a sequence of updates winning the locking
1368 battle against the aggregator. Only if an aggregator has managed to
1369 merge data over to the next level, truncating can be allowed. DONE with
1370 accessor merged.
1372 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1374 * We should probably guarantee that no duplicates enter the aggregator
1375 array.
1377 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1379 * To get merge operation faster would need a good benchmark test. What
1380 02 spits out isn't reliable enough and is dominated by many other
1381 things. Between
1383 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1384 Author: Andreas J. Koenig <k@k75.(none)>
1385 Date: Sat Aug 2 07:58:04 2008 +0200
1387 and
1389 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1390 Author: Andreas J. Koenig <k@k75.(none)>
1391 Date: Sat Aug 2 11:40:29 2008 +0200
1393 there was a lot of trying but the effect is hardly measurable with
1394 current tests.
1396 * overhead of connecting seems high. When setting
1397 max_files_per_connection to 1 we see that.
1399 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1401 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1403 25138.2215600014/3600 = 6.98283932222261
1405 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1406 be a bug in the aggregator. FIXED
1408 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1410 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1411 we have it, can we mirror it, mirror it!
1413 I fear this needs a new class which might be called
1414 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1415 and treat them as an entity. I realize that a single recentfile may be
1416 sufficient for certain tasks and that it is handy for the low level
1417 programmer but it is not nice to use. If there is a delete in the 1h
1418 file then the 6h file still contains it. Seekers of the best information
1419 need to combine at least some of the recentfiles most of the time.
1421 There is the place for the Z loop!
1423 But the combination is something to collect in a database, isn't it. Did
1424 csync2 just harrumph?
1426 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1428 * it just occurred to me that hosts in the same mirroring pool could
1429 help out each other even without rewriting the recentfile. Just fetch
1430 the stuff to mirror from several places, bingo. But that's something
1431 that should rather live in a separate package or in rsync directly.
1433 * cronjobs are unsuited because with ntp they would all come at the full
1434 minute and disturb each other. Besides that I'd hate to have a backbone
1435 with more than a few seconds latency.
1437 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1439 * a second rsync server with access control for PAUSE. Port? 873 is the
1440 standard port, let's take 8873.
1442 * if there were a filesystem based on this, it would have a slow access
1443 to inexistent files. It would probably provide wrong readdir (only based
1444 on current content) or also a slow one (based on a recentfile written
1445 after the call). But it would provide fast access to existing files. Or
1446 one would deliberately allow slightly blurred answers based on some
1447 sqlite reflection of the recentfiles.
1449 * todo: write a variant of mirror() that combines two or more
1450 recentfiles and treats them like one
1452 * todo: signal handler to remove the tempfile
1454 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1456 * now that we have the symlink I forgot how it should be used in
1457 practice.
1459 * the z loop: add missing files to Z file. Just append them (instead of
1460 prepending). So one guy prepends something from the Y file from time to
1461 time and another guy appends something rather frequently. Collecting
1462 pond. When Y merges into Z, things get epoch and the collecting pond
1463 gets smaller. What exactly are "missing files"?
1465 take note of current epoch of the alpha file, let's call it the
1466 recent-ts
1468 find all files on disk
1470 remove all files registered in the recentworld up to recent-ts
1472 remove all files that have been deleted after recent-ts according to
1473 recentworld
1475 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1477 * rersyncrecent might be a cronjob with a (locked) state file which
1478 contains things like after and maybe last z sync or such?
1480 rrr-mirror might be an alternative name but how would we justify the
1481 three Rs when there is no Re-Rsync-Recent?
1483 With the --loop parameter it is an endless loop, without it is no loop.
1484 At least this is simple.
1486 * todo: new accssor z-interval specifies how often the Z file is updated
1487 against the filesystem. We probably want no epoch stamp on these
1488 entries. And we want to be able to filter the entries (e.g. no
1489 by-modules and by-category tree)
1491 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1493 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1494 into Z. We do not need the other files filled up with filesystem
1495 contents though.
1497 * need interface to query for a file in order to NOT call update on
1498 PAUSE a second time within a short time.
1500 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1502 * recommended update interval? Makes no sense, is different for
1503 different users.
1505 * Moosify
1507 Local Variables:
1508 mode: change-log
1509 change-log-default-name: "Todo"
1510 tab-width: 2
1511 left-margin: 2
1512 End: