| blob | 2cb00df7f74bdcb0326892297eabd5679abc3b12 |
1 2011-02-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * lockdirectory expiration? server died and blocked fsck for so long.
5 * At the moment rrr-init is not needed anymore, rrr-server can be
6 started and followed by rrr-fsck. Still stupid but not the main
7 showstopper.
9 2011-02-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
11 * IN_Q_OVERFLOW needs to trigger an fsck, likewise the entry into the
12 server.
14 * speed: a large directory remove that is done by the kernel in 1
15 seconds triggers minutes of bookkeeping (07:56:42 - 08:01:10 for ~3000
16 files, all due to Schlemiehl again. Want to do some collecting of
17 pending ops before actually writing to the RECENT files. Lock!
19 2011-01-30 Andreas J. Koenig <andk@cpan.org>
21 * rrr-init and rrr-server need to be used such:
23 init
24 server
25 init again
27 Stupid but no showstopper ATM.
29 2010-10-26 Andreas J. Koenig <andk@cpan.org>
31 * forking slave is done, needs more testing.
33 2010-07-06 Andreas J. Koenig <andk@cpan.org>
35 * The two next most important goals are to get a useable rrr-server and
36 a forking slave implementation that uses virtually no memory after the
37 first pass through. Does it need the status file?
39 * rrr-init now being fast we can now remove the alpha alerts. Done.
41 2010-06-16 Andreas J. Koenig <andk@cpan.org>
43 * rrr-init seems to be sssslllloooowwww. At 3:49 I started it for the
44 video tree. After twelve minutes the file has reached a size of 500k. It
45 is locking and rewriting the YAML after every file. Wasn't there an
46 unlocked variant? For "init" one would not expect locked correctness and
47 a rewrite after every item. Given that this tends to become slower with
48 every file it is really unacceptable.
50 Reminds me of rrr-dirtyupdate (the former name of rrr-update). It seems
51 the code was not dirty enough and that was the reason why it was so
52 slow. And rrr-update does not have a --dirty switch, so it looks like we
53 have no way to force dirty and fast operation yet but we need it for
54 init.
56 Irritating the bootstrapping race: the order {(1) fill initial
57 array, (2) start server} leaves the need for an fsck to cover the time
58 between 1 and 2. Given that regular fsck is needed in any case the price
59 is not high but still doesn't give a warm feeling. Maybe starting the
60 server will by itself do an fsck?
62 2010-06-15 Andreas J. Koenig <andk@cpan.org>
64 * remove the wild alpha alerts and say, it is alpha but parts of it are
65 already working very well or so.
67 2009-12-10 Andreas J. Koenig <andk@cpan.org>
69 * Prominent request for statusfile to safe memory consumption. I'd like
70 to work backwards: get a verification program that verifies how well in
71 sync two nodes are without using frmr itself. Pretend that the forking
72 agent already works and can be interrupted with ^C. Then let it die at
73 random and restart at random and see the memory consumption and the
74 failure rate over time. Compare it with the curve we have now.
76 Of course we also invented the status file as a sharp debugging tool.
77 I'd like to see it in action. See also: F:R:M:Recent::thaw().
79 2009-06-22 Andreas J. Koenig <andk@cpan.org>
81 * Todo: rrr-init to generate some index files so that rrr-server can be
82 started.
84 2009-05-02 Andreas J. Koenig <andk@cpan.org>
86 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
87 so we can use it for normal update with a single file. And add a cronjob
88 that runs rrr-fsck so we get a reminder when something is not working
89 as it should.
91 2009-04-27 Andreas J. Koenig <andk@cpan.org>
93 * Todo: I still see more link_stat errors than expected. I come back to
94 the former suspicion that we do not do necessary bookkeeping on deleted
95 files so that when we reach the "new" event on a meanwhile deleted file
96 we try to mirror it although we could know better. It's just noise, no
97 harm but noise is irritating.
99 * Todo: more experimenting with runstatusfile to get it official
101 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
102 parse the error.
104 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
106 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
108 And that ignore_link_stat_error is propagated in sparse_clone.
110 DONE
112 For the record: the bug was that ignore_link_stat_error was lost during
113 sparse_clone. And then the second bug was that it did not default to
114 true. Both are now fixed.
116 2009-04-25 Andreas J. Koenig <andk@cpan.org>
118 * Todo: fill rrr program with life
120 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
122 * Minor bug: I see complaints about files not existing like
123 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
124 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
126 In the RECENT files these are delete events but of course also new
127 events in older recent files.
129 I think we do not properly remember deletes when we skip-deletes and so
130 do not filter them out when we later see the "new" event.
132 Yes, but that what ignore_link_stat_error is for, we know about this
133 race condition. We must make ignore_link_stat_error default to true.
135 DONE
137 * bug in Done.pm failing to merge two fields into one when a third is
138 present. FIXED and accompanied with new test.
140 2009-04-24 Andreas J. Koenig <andk@cpan.org>
142 * Todo: keep the index clean and run some kind of nondestructive fsck 4
143 times a day.
145 * who is our backbone?
147 http://cpan.cpantesters.org/authors/02STAMP # barbie
148 http://cpan.solfo.com/authors/02STAMP # abh
149 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
151 2009-04-17 Andreas J. Koenig <andk@cpan.org>
153 * the changelog helper I included in the Makefile release memo does not
154 show tags and is much less useful than I thought. gitk is probably much
155 more convenient.
157 2009-04-16 Andreas J. Koenig <andk@cpan.org>
159 * ABH writes:
161 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
162 #### Ydr_.yaml ... DONE
163 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
164 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
165 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
166 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
167 #### gone already at cpan-pause.pl line 0
169 Running without skip-deletes now but cannot reproduce.
171 2009-04-15 Andreas J. Koenig <andk@cpan.org>
173 * consider state directory such we can restart after a ^C or come again
174 from a cronjob.
176 Reuse _runstatusfile: write atomically, see how we can use the dump to
177 start running again. Do not leak and remove globs before writing. Do not
178 only dump one rf, dump the whole complex.
180 * consider the effect of ^C. Is it safe?
182 * want to use Perl::Version to up the versions on every release. Only if
183 things have changed, of course. Does this actually work?
185 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
186 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
188 Must do the bookkeeping myself when this should happen, maybe based on git?
190 I think I prefer the same version number for all pm files. But one also
191 needs a routine for setting them. Ahh:
193 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
195 NEEDSMOREWORK
197 * why indexer wrong?
199 cpan[2]> m /Mirror::Recent/
200 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
201 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
202 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
203 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
204 4 items found
206 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
208 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
210 I just tried MM 6.5102 and it again produces only a provides for Recent,
211 not for the other modules.
213 My own fault. FIXED in the Makefile.PL.
215 2009-04-13 Andreas J. Koenig <andk@cpan.org>
217 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
219 * is it true (as stated at
220 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
221 that rsync can lead to errors when upstream changes before client sync
222 has completed? Can we see the error when we run this on upstream:
224 % perl -e '
225 use Time::HiRes qw(time);
226 while (){
227 open my $fh, ">", "changingfile.txt.new" or die;
228 print $fh (time."\n") x 1000000;
229 close $fh;
230 rename "changingfile.txt.new", "changingfile.txt" or die;
231 }
232 '
234 And on the receiving end:
236 % while true; do
237 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
238 sleep 1;
239 done
241 I cannot get it to fail. Apparently it is sufficient when upstream
242 always writes atomically (which of course is mandatory)?
244 SENT to mailing list Apr 28.
246 * from ABH:
248 > https://fedorahosted.org/InstantMirror/
249 > https://www.redhat.com/mailman/listinfo/instantmirror-list
250 > irc.freenode.net channel #instantmirror
252 A cool name for a project. Inspires me to write a few sentences about
253 bittorrent's role in the grand picture.
255 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
256 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
258 2009-04-12 Andreas J. Koenig <andk@cpan.org>
260 * Interesting last minute bug during real download testing: the output
261 isn't as pretty anymore, it seems that more work is being done than
262 needed.
264 Introducing a highlevel _alluptodate method helped a lot in debugging
265 this.
267 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
268 different on 1h file than on the other files. This lead to frequent
269 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
270 issue.
272 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
274 * Todo: try if we can be faster with a native float package. I'm really
275 glad to have a machine exposing perl floating point bugs, it forced me
276 to retract this item from the changes file: "several speedups in the
277 fakebigfloat code" before I get cpan testers fails.
279 2009-04-11 Andreas J. Koenig <andk@cpan.org>
281 * Bug: when the dirtymark changes upstream, then the downstream server
282 notices it immediately and mirrors the "1h" file but then leaves the
283 rmirror loop and continues with "6h" on the next round through. This
284 behaviour goes on until it reaches the Z file. Because we eat the Z file
285 piecemeal we leave the loop after a while but after that everybody seems
286 to have forgotten that there is some work left to be done. Bad, bad,
287 bad.
289 ETOOCONFUSING better name?
291 Recentfile::get_remote_recentfile_as_tempfile OK
292 Recentfile::resolve_recentfilename split_rfilename DONE
294 Recent::_principal_recentfile_object _principal_recentfile_fromremote
295 Recent::principal_recentfile OK
296 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
298 some void, some not, some not void but called in void context.
300 2009-04-10 Andreas J. Koenig <andk@cpan.org>
302 * Bug? When the dirtymark changes then the second tier hosts quickly
303 reset their DONE state and restart mirroring. But then they mirror
304 potentially outdated index files with inconsistent dirtymark because the
305 first tier box may be a bit behind on the large recentfiles.
307 This means we need a brake when iterating over the recentfiles that
308 refuses to use a recentfile with the wrong dirtymark.
310 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
312 * install the bin/* files and talk about rrr-overview somewhere.
314 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
315 created outside the target tree.
317 * Barbie asks for better logging capabilities
319 2009-03-29 Andreas J. Koenig <andk@cpan.org>
321 * known bug: cannot configure the Recentfile objects to keep delete
322 events. How to call the accessor for that? keep_delete_objects_forever?
324 Sounds acceptable. FIXED (but untested).
326 * known bug: keeps temporary index files lying around.
328 2009-03-24 Andreas J. Koenig <andk@cpan.org>
330 * Some equivalent for
332 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
334 ? DONE
336 2009-03-22 Andreas J. Koenig <andk@cpan.org>
338 * k81 is again client of k75 with different parameters.
340 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
341 about it all the time and nobody reacts accordingly; (2) lots of temp
342 files get created and not removed; culprit the new call to
343 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
344 caller has to remove the tempfile after use.
346 Need the drawing board.
348 PARTLY FIXED: retracting the idea to call
349 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
350 and unseed such that uptodate now means "we have mirrored this
351 recentfile's bunch of files" and seeded means "somebody believes the
352 index file for this rf needs be refreshed". This seems to work now.
354 Still collecting temp files that nobody cares to depollute.
356 2009-03-21 Andreas J. Koenig <andk@cpan.org>
358 * Bug?: should it be harder than it is atm to set the timestamp to the
359 future? YES, FIXED
361 * bug with native integers:
363 -
364 epoch: 997872011
365 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
366 type: new
367 -
368 epoch: 1195248431
369 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
370 type: new
372 Native integer broke when native math was turned off. FIXED
374 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
375 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
376 in this area.
378 Yes, records out of order:
380 447003 -
381 447004 epoch: 995885533
382 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
383 447006 type: new
384 447007 -
385 447008 epoch: 995890358
386 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
387 447010 type: new
388 447011 -
389 447012 epoch: 995892221
390 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
391 447014 type: new
393 FIXED with sanity check and later with the integer fix.
395 * Bug: want the index files in a .recent directory
397 * Bug: lots of dot files are not deleted in time
399 * possible test case: can a delete change the timestamp? This would
400 probably break the order of events.
402 2009-03-20 Andreas J. Koenig <andk@cpan.org>
404 * 1233701831.34486 what's so special about this number/string? It
405 belongs to
406 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
407 atm lives in Y,Q, and Z.
409 It is the first entry after id/--skip-locking which has timestamp
410 1234164228.11325 which represents a file that doesn't exist anymore.
412 * another bug:
414 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
415 _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
416 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
417 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
418 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
419 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
420 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
421 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
422 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
423 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
424 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
425 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
426 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
428 FIXED, it was the "--skip-locking" file where manual intervention was participating
430 * bug on the mirroring slave: when the dirtymark gets increased we
431 probably do not reset the done intervals. The mirrorer stays within
432 tight bounds where it tries to sync with upstream and never seems to
433 finish. In the debugging state file I see lots of identical intervals
434 that do not get collapsed. When I restart the mirrorer it dies with:
436 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
437 _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
438 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
439 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
440 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
441 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
442 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
443 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
444 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
445 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
447 and debugging stands at
449 At: before
450 Brfinterval: Z
451 Ci: 227
452 Dre+1:
453 epoch: 1237400802.5789
454 path: id/J/JO/JOHND/CHECKSUMS
455 type: new
456 Dre-0:
457 epoch: 1237400807.97514
458 path: id/M/MI/MIYAGAWA/CHECKSUMS
459 type: new
460 Dre-1:
461 epoch: 1237400817.94363
462 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
463 type: new
464 Eintervals:
465 -
466 - 900644040
467 - 900644040
468 - []
470 and it is reproducable.
472 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
473 pause has a fresh one.
475 FIXED, it was the third anded term in each of the ifs in the IV block in
476 _register_one: with that we make sure that we do not stamp on valuable
477 interval data.
479 2009-03-17 Andreas J. Koenig <andk@cpan.org>
481 * done: verified the existence of the floating point bug in bleadperl
482 and verified that switching from YAML::Syck to YAML::XS does not resolve
483 it.
485 BTW, the switch was doable with
487 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
489 and should be considered as a separate TODO
491 * todo: integrate a dirty update with two aggregate calls before
492 unlocking for frictionless dirtying DONE
494 * todo: start the second rsync daemon on pause DONE
496 * todo: move index files to .recent: this cannot simply be done by
497 setting filenameroot to .recent/RECENT. Other parts of the modules rely
498 on the fact that dirname(recentfile) is the root of the mirrored tree.
500 2009-03-16 Andreas J. Koenig <andk@cpan.org>
502 * What was the resolution of the mirror.pl delete hook bug? Do we call
503 the delete hook when pause removes a file from MUIR?
505 * Today on pause: Updating 2a13fba..29f284d and installing it for
506 /usr/local/perl-5.10.0{,-RC2}
508 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
509 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
511 2009-03-15 Andreas J. Koenig <andk@cpan.org>
513 * currently recent_events has the side effect of setting dirtymark
514 because it forces a read on the file. That should be transparent, so
515 that the dirtymark call always forces a cache-able(?) read.
517 * The bug below is -- after a lot of trying -- not reproducible on a
518 small script, only in the large test script. The closest to the output
519 below was:
521 #!perl
522 use strict;
523 use Devel::Peek;
524 my $x = "01237123229.8814";
525 my($l,$r);
526 for ($l,$r) {
527 $_ = "x"x34;
528 }
529 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
530 $r = int $r;
531 $l = "1237123231.22458";
532 $r = "1237123231.22458";
533 1 if $l/1.1;
534 Devel::Peek::Dump $l;
535 Devel::Peek::Dump $r;
536 Devel::Peek::Dump $x = $l <=> $r;
537 die "BROKE" if $x;
538 __END__
540 The checked in state at c404a85 fails the test with my
541 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
542 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
543 in the test script always consistent.
545 * Todo: write a test that inserts a second dirty file with an already
546 existing timestamp. DONE
548 * Bug in perl 5.10 on my 64bit box:
550 DB<98> Devel::Peek::Dump $l
551 SV = PVMG(0x19e0450) at 0x142a550
552 REFCNT = 2
553 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
554 IV = 0
555 NV = 1237123231.22458
556 PV = 0x194ce70 "1237123231.22458"\0
557 CUR = 16
558 LEN = 40
560 DB<99> Devel::Peek::Dump $r
561 SV = PVMG(0x19e0240) at 0x142a3e8
562 REFCNT = 2
563 FLAGS = (PADMY,POK,pPOK)
564 IV = 1237123229
565 NV = 1237123229.8814
566 PV = 0x19ff900 "1237123231.22458"\0
567 CUR = 16
568 LEN = 40
570 DB<100> Devel::Peek::Dump $l <=> $r
571 SV = IV(0x19ea6e8) at 0x19ea6f0
572 REFCNT = 1
573 FLAGS = (PADTMP,IOK,pIOK)
574 IV = -1
576 DB<101> Devel::Peek::Dump $l
577 SV = PVMG(0x19e0450) at 0x142a550
578 REFCNT = 2
579 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
580 IV = 0
581 NV = 1237123231.22458
582 PV = 0x194ce70 "1237123231.22458"\0
583 CUR = 16
584 LEN = 40
586 DB<102> Devel::Peek::Dump $r
587 SV = PVMG(0x19e0240) at 0x142a3e8
588 REFCNT = 2
589 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
590 IV = 1237123231
591 NV = 1237123231.22458
592 PV = 0x19ff900 "1237123231.22458"\0
593 CUR = 16
594 LEN = 40
596 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
598 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
600 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
602 * $obj->merge ($other) needs to learn about equal epoch which may happen
603 since dirty_epoch intruded.
605 * Wontfix anytime soon: I think we currently do not support mkdir. Only
606 files!
608 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
610 * Todo: continue working on update(...,$dirty_epoch). It must be
611 followed by a fast_aggregate! DONE
613 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
615 * maybe we need a closest_entry or fitting_interval or something like
616 that. We want to merge an event into the middle of some recentfile.
617 First we do not know which file, then we do not know where to lock,
618 where to enter the new item, when and where to correct the dirtymark.
620 So my thought is we should first find which file.
622 Another part of my brain answers: what would happen if we would enter
623 the new file into the smallest file just like an ordinary new event,
624 just as an old event?
626 (1) we would write a duplicate timestamp? No, this would be easy to
627 avoid
629 (2) we would make the file large quickly? Yes, but so what? We are
630 changing the dirtymark, so are willing to disturb the downstream hosts.
632 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
634 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
636 leak!
638 https://rt.cpan.org/Ticket/Display.html?id=41199
640 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
642 Strangest output being HAKANARDO who managed to upload
644 Here is a better oneliner that includes also the first line of each
645 finding:
647 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
649 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
651 * memory leak in the syncher? It currently weighs 100M.
653 Update 2008-11-02:
655 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
658 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
660 * lookup by epoch and by path and use this ability on the pause to never
661 again register a file twice that doesn't need it. Let's call it
662 contains().
664 * after the dirtymark is done: fill up recentfiles with fake (historic)
665 entries; fill up with individual corrections; algorithm maybe to be done
666 with bigfloat so that we can always place something in the middle
667 between two entries. Before we must switch to bigfloat we could try to
668 use Data::Float::nextup to get the.
670 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
671 directly.
673 * dirtymark now lives in Recentfile, needs to be used in rmirror.
675 * find out why the downloader died after a couple of hours without a net
676 connection. Write a test that survives the not-existence of the other
677 end forever.
679 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
681 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
682 it makes less sense with tight coupling for secondary files. But asking
683 the server what the current epoch is might be cheaper on HTTP than on
684 rsync. (Needs to be evaluated)
686 * remove the 0.00 from the verbose overview in the Merged column in the
687 Z row. DONE
689 * write tests that expose the problems of the last few days: cascading
690 client/server roles, tight coupling for secondary RFs, deletes after
691 copies.
693 * Some day we might want to have policy options for the slave:
694 tight/loose/no coupling with upstream for secondary RFs. tight is what
695 we have now. loose would wait until a gap occurs that can be closed.
697 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
699 * revisit all $rfs->[$i+1] places if they now make sense still
701 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
703 * another bug is the fact that the mirror command deletes files before
704 it unhides the index file, thus confusing downstream slaves. We must not
705 delete before unhiding and must delete after unhiding. FIXED.
707 * new complication about the slave that is playing a server role.
708 Currently we mirror from newest to oldest with a hidden temporary file
709 as index. And when one file is finished, we unhide the index file.
710 Imagine the cascading server/slave is dead for a day. It then starts
711 mirroring again with the freshest thing and unhides the freshest index
712 file when it has worked through it. In that moment it exposes a time
713 hole. Because it now works on the second recentfile which is still
714 hidden.
716 We currently do nothing special to converge after such a drop out. At
717 least not intentionally and robustly and thought through.
719 The algorithm we use to seed the next file needs quite a lot of more
720 robustness than it currently has. Something to do with looking at the
721 merged element of the next rf and when it has dropped off, we seed
722 immediately. And if it ramains dropped off, we seed again, of course.
724 Nope, looking from smaller to larger RFS we look at the merged element
725 of this RF and at the minmax/max element of the next RF. If that
726 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
727 seeding it.
729 And we need a public accessor seed and unseed or seeded. But not the mix
730 of public and private stuff that then is used behind the back.
732 And then the secondary* stuff must go.
734 And we must understand what the impact is on the DONE system. Can it go
735 unnoticed that there was a hole? And could the DONE system have decided
736 the hole is covered? This should be testable with three directories where
737 the middle stops working for a while. Done->merge is suspicious, we must
738 stop it from merging non-conflatable neighbors due to broken continuity.
740 FIXED
742 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
744 * Slaven suggests to have the current epoch or the whole current
745 recentfile available from the HTTP server and take it away with
746 keepalive. This direction goes the granularity down to subseconds.
748 We might want to rewrite everything to factor out transport and allow
749 the whole thing to run via HTTP.
751 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
753 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
754 symlink?
756 * smoker on k81 fetching from k75 to verify cascading works. See
757 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
759 * maybe the loop should wait for CHECKSUMS file after every upload. And
760 CPAN.pm needs to deal with timestamps in the future.
762 * do not forget the dirtymark!
764 Text: have a new flag on recentfiles with the meaning: if this
765 changes, you're required to run a full rsync over all the files. The
766 reason why we set it would probably be: some foul happened. we injected
767 files in arbitrary places or didn't inject them although they changed.
768 The content of the flag? Timestamp? The relation between the
769 recentfiles would have to be inheritance from the principal, because any
770 out of band changes would soon later propagate to the next recentfile.
772 By upping the flag often one can easily ruin the slaves.
774 last out of band change? dirtymark?
776 Anyway, this implies that we read a potentially existing recentfile
777 before we write one.
779 And it implies that we have an eventloop that keeps us busy in 2-3
780 cycles, one for current stuff (tight loop) and one for the recentfiles
781 (cascade when principal has changed), one for the old stuff after a
782 dirtymark change.
784 And it implies that the out-of-band change in any of the recentfiles
785 must have a lock on the principal file and there is the place to set the
786 dirtymark.
788 * start a FAQ, especially quick start guide questions. Also to aid those
789 problematic areas where we have no good solution, like the "links"
790 option to rsync.
792 * wish feedback when we are slow.
794 * reduce mccabe
796 * Remove a few DEBUG statements.
798 * The multiple-rrr way of doing things needs a new option to rmirror,
799 like piecemeal or so. Not urgent because after the first pass through,
800 things run smoothely. It's only ugly during the first pass.
802 * I have the suspicion that the code is broken that decides if the
803 neighboring RF needs to be seeded. I fear when too much time has gone
804 between two calls (in our case more than one hour), it would not seed
805 the neighbor. Of course this will never be noticed, so we need a good
806 test for it.
808 * local/localroot confusion: I currently pass both options but one must
809 do.
811 * accounts for early birds on PAUSE rsync daemon.
813 * hardcoded 20 seconds
815 * who mirrors the index? DOING now.
817 * which CPAN mirrors offer rsync?
819 * visit all XXX, visit all _float places
821 * rename the pathdb stuff, it's too confusing. No idea how.
823 * rrr-inotify, backpan, rrr-register
825 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
827 * current bugs: the pathdb seems to get no reset, the seeding of the
828 secondaryttl stuff seems not to have an effect. Have helped myself with
829 a rand(10), need to fix this back. So not checked in. Does the rand
830 thing even help?
832 The rand thing helps. The secondaryttl stuff was in the wrong line,
833 fixed now.
835 The pathdb stuff was because I called either _pathdb or __pathdb on the
836 wrong object. FIXED now.
838 * It's not so beautiful if we never fetch the recentfiles that are not
839 the principal, even if this is correct behaviour. We really do not need
840 them after we have fetched the whole content.
842 OK, we want a switch for that: secondaryttl DONE
844 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
846 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
847 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
848 it lets 40 follow. FIXED
850 * See that the long running process really only updates the principal
851 file unless it has missed a timespan during which something happened. If
852 nothing happened, it must notice even when it misses the timespan. DONE
854 * we must throw away the pathdb when we have reached the end of Z. From
855 that moment we can have a very small pathdb because the only reason for
856 a pathdb is that we know to ignore old records in old files. We won't
857 need this pathdb again before the next full pass over the data is
858 necessary and then we will rebuild it as we go along. DONE
860 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
862 * I think, Done::register_one is doing wrong in that it does not
863 conflate neighboring pieces. The covered() method cannot do this because
864 it has no recent_events array at hand. But register_one has it and could
865 do it and for some reason misses to do it (sometimes).
867 This means that the three tests I just wrote can probably not survive
868 because they test with an already broken Done structure.
870 The art now is to detect how it happens, then to reproduce, then write a
871 test, then fix it.
873 So from the logfile this is what happens: we have a good interval with
874 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
875 top of it. Locally we now mirror F2 and open a new done interval for it.
876 Then we mirror F1 but this time with the timestamp T1b. And when we then
877 try to close the gap, we do not find T1 but instead something older. We
878 should gladly accept this older piece and this would fix this bug.
880 FIXED
882 * bug to fix: when the 1h file changes while rmirror is running, we do
883 correctly sync the new files but never switch to the 6h file but rather
884 stay in a rather quick loop that fetches the 1h file again and again.
886 Is it possible that we initialize a new object? Or does
887 get_remote_recentfile_as_tempfile overwrite something in myself?
889 Want a new option: _runstatusfile => $file which frequently dumps the
890 state of all recentfiles to a file.
892 FIXED
894 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
896 * Todo: now teach update to verify the timestamp is about to write
897 against the previous and use _increase_a_bit if it doesn't comply with
898 strict monotony. DONE
900 * The problem of rounding. So far perl's default precision was
901 sufficient. One day it won't be. FakeFloat has an easy job when it is
902 only reading and other machines have written correctly. But when we want
903 to write a floating point number that is a bit larger than the other
904 one, then we need our own idea of precision.
906 Slaven said: just append a "1". This might be going towards the end of
907 usability too quickly. I'd like something that actually uses the decimal
908 system. Well, appending a 1 also does this but...
910 E.g. we have 1.0. nextup on this architecture is starting with
911 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
912 taken the 1.0000000000000003 and the next user comes and the time tells
913 him 1.0 again. He has to beat my number without stepping over the
914 nextup. This is much less space than I had when I chose 1,2,3.
916 What is also irritating is that nextup is architecture dependent. The
917 128 bit guy must choose very long numbers to fit in between whereas the
918 other one with 16 bit uses larger steps. But then the algorithm is the
919 same for both, so that would be a nice thing.
921 I see two situation where we need this. One is when Time::HiRes returns
922 us a value that is <= the last entry in our recentfile. In this case
923 (let's call it the end-case) we must fill the region between that number
924 and the next higher native floating point number. The other is when we
925 inject an old file into an old recentfile (we would then also set a new
926 dirtymark). We find the integer value already taken and need a slightly
927 different one (let's call it the middle-case). The difference between
928 the two situations is that the next user will want to find something
929 higher than my number in the end-case and something lower than my number
930 in the middle case.
932 So I suggest we give the function both a value and an upper bound and it
933 calculates us a primitive middle. The upper bound in the middle-case is
934 the next integer. The upper bound on the end-case is the nextup floating
935 point number. But the latter poses another problem: if we have occupied
936 the middle m between x and nextup(x), then the nextup(m) will probably
937 not be the same as nextup(x) because some rounding will take place
938 before the nextup is calculated and when the rounding reaches the
939 nextup(x), we will end up at nextup(nextup(x)).
941 So we really need to consider the nextup and the nextdown from there and
942 then the middle and that's the number we may approach asymptotically.
943 Ugly. But DONE.
945 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
947 * consider deprecating the use of RECENT.recent as a symlink. It turns
948 out to need extra hoops with the rsync options and just isn't worth it.
949 Or maybe these extra hoops are needed anyway for the rest of the tree?
950 Nope, can't be the case because not all filesystems support symlinks.
952 But before doing the large step, I'll deprecate the call of
953 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
954 only introduced to resolve RECENT.recent and complicates the routine far
955 beyond what it deserves.
957 DONE. Won't deprecate RECENT.recent, just moved its handling to the
958 supervisor.
960 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
962 * I think it's a bug that the rsync_option links must be set to true in
963 order to support RECENT.recent and that nobody cares to set it
964 automatically. Similar for ignore_link_stat_errors.
966 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
968 * Todo: find all todos together and make a plan what is missing for a
969 release.
971 - verifytree or something like that. fsck maybe.
973 - rersyncrecent, the script itself? What it do?
975 - a way to only mirror the recentfiles without mirroring the whole
976 remote system such that people can decide to mirror only partially see
977 also 2008-08-30. .shadow-xxx directory? this also needed for a
978 filesystem that is still incomplete and might need the mirrorfiles for
979 lookup(?)
981 - long living objects that mirror again and again. Inject something
982 into ta, see how it goes over to tb.
984 - how do we continue filling up the DONE system when we use an object
985 for the second time? "fully covered" and "uptodate" or new terminology.
987 - overview called on the wrong file should be understandable
989 - the meta data field that must change when we fake something up so that
990 the downstream people know they have to re-fetch everything.
992 - how tolerant are we against missing files upstream? how do we keep
993 track? there are legitimate cases where we did read upstream index right
994 before a file got deleted there and then find that file as new and want
995 it. There are other cases that are not self healing and must be tracked
996 and bugreported.
998 - how, exactly, do we have to deal with deletes? With rsync errors?
1000 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
1001 authors) failed: No such file or directory (2)
1003 The file above is a delete in 1h and a new in file 1M and the
1004 delete in the locally running rmirror did not get propagated to the 1M
1005 object. Bug. And the consequence is a standstill.
1007 It seems that a slave that works with a file below the principal needs
1008 to merge things all the way up to get rid of later deletes. Or keep
1009 track of all deletes and skip them later. So we need a trackdeletes.pm
1010 similar to the done.pm?
1012 see also 2008-08-20 about spurious deletes that really have no add
1013 counterpart and yet they are not wrong.
1015 - consider the effect when resyncing the recentfile takes longer than
1016 the time per loop. Then we never rsync any file. We need to diagnose
1017 that and force an increase of that loop time. But when we later are fast
1018 enough again because the net has recovered, then we need to switch back
1019 to original parameters. ERm, no, it's enough to keep syncing at least
1020 one file before refetching an index file.
1022 - remember to verify that no temp files are left lying around and the
1023 signal handler
1025 - status file for not long running jobs that want to track upstream with
1026 a, say, cronjob.
1028 - revisit all XXX _float areas and study Sub::Exporter DONE
1030 - persistent DB even though we just said we do not need it. Just for
1031 extended capabilities and time savings when, for example, upstream
1032 announces a reset and we get new recentfiles and could then limit
1033 ourselves to a subset of files (those that have a changed epoch) in a
1034 first pass and would only then do the loop to verify the rest. Or
1035 something.
1037 * Todo: aggregate files should know their feed and finding the principal
1038 should be done stepwise. (?)
1040 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
1041 left locks around. DONE
1043 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1045 * maybe extend the _overview so that it always says if and where the
1046 last file is in the next file and where the next event in the next rf
1047 would lie. No, don't like this anymore. REJECT
1049 * take the two new redundant tests out again, only the third must
1050 survive. DONE
1052 * Todo: add a sanity check if the merged structure is really pointing to
1053 a different rf and that this different rf is larger. DONE
1055 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1057 * now test, if they are overlapping. And test if there is a file in the
1058 next rf that would fit into this rf's interval.
1060 1h 1222324012.8474 1222322541.7963 0.4086
1061 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
1062 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
1063 1W 1222313218.3626 1221708477.5829 167.9835
1065 I suspect that somebody writes a merged timestamp without having merged
1066 and then somebody else relies on it.
1068 If aggregate is running, the intervals must not be extravagated, if it
1069 is not running, there must not be bounds, the total number of events in
1070 the system must be counted and must be controlled throughout the tests.
1071 That the test required the additional update was probably nonsense,
1072 because aggregate can cut pieces too. FIXED & DONE
1074 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1076 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1077 changed. FIXED
1079 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1081 * Most apparent bug at the moment is that the recentfiles are fetched
1082 too often. Only the principal should be fetched and if it has not
1083 changed, the others should not be refetched. ATM I must admit that I'm
1084 happy that we refetch more often than needed because I can more easily
1085 fix bugs while the thing is running.
1087 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1088 done but the done system does not know about it. The reason for the
1089 failure is not known and we never reach the status uptodate because of
1090 this. We must get over it.
1092 Later it turns out that the origin server had a bug somewhere.
1093 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1094 array of the recentfile one position above. The bug was my own.
1096 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1098 * There is the race condition where the server does a delete and the
1099 slave does not yet know and then tries to download it because he sees
1100 the new. So for this time window we must be more tolerant against
1101 failure. If we cannot download a file, we should just skip it and should
1102 not retry immediately. The whole system should discover the lost thing
1103 later. Keeping track with the DONE system should really be a no brainer.
1105 But there is something more: the whole filesystem is a database and the
1106 recentfiles are one possible representation of it. It's a pretty useful
1107 representation I think that's why I have implemented something around
1108 it. But for strictly local operation it has little value. For local
1109 operation we would much rather have a database. So we would enter every
1110 recentfile reading and every rsync operation and for every file the last
1111 state change and what it leads to. Then we would always ignore older
1112 records without the efforts involved with recentfiles.
1114 The database would have: path,recentepoch,rsyncedon,deletedon
1116 Oh well, not yet clear where this leads to.
1118 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1120 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1121 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1123 * statusfile, maybe only on demand, alone to have a sharp debugging
1124 tool. It is locked and all recentfiles dump themselves into it and we
1125 can build a viewer that lets us know where we stand and what's inside.
1127 * remember: only the principal recentfile needs expiration, all others
1128 shall be expired by principal if it discovers that something has move
1129 upstream.
1131 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1133 * Always check if we stringify to a higher value than in the entry
1134 before. DONE
1136 * And in covered make an additional check if we would be able to see a
1137 numerical difference between the two numbers and if we can't then switch
1138 to a different, more expensive algorithm. Do not want to be caught by
1139 floating surprises. DONE
1141 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1143 * caching has several aspects here: we can cache the interval of the
1144 recentfile which only will change when the mtime of the file changes. We
1145 must re-mirror the recentfile when its ttl has expired. Does have_read
1146 tell you anything? It counts nothing at all. Only the mtime is
1147 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1148 thing the have_mirrored because it is unrelated to the mtime.
1150 * Robustness of floating point calculations! I always thought that the
1151 string calculated by the origin server for the floating representation
1152 of the epoch time is just a string. When we convert it to a number and
1153 later back to a string, the other computer might come to a different
1154 conclusion. This must not happen, we want to preserve it under any
1155 circumstances. I will have to write tests with overlong sequences that
1156 get lost in arithmetic and must see if all still works well. DONE
1158 But one fragile point remains: if one host considers a>b and the other
1159 one considers them == but no eq. To prevent this, we must probably do
1160 some extra homework. DONE
1162 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1164 * the concept of tracking DONE needs an object per recentfile that has
1165 something like these methods:
1167 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1169 covered() register() covered()
1171 The unclear thing is how we translate points in time into intervals. We
1172 could pass a reference to the current recent_events array when running
1173 we_have(xxx) and let the DONE object iterate over it such that it only
1174 has to store a list of intervals that can melt into each other. Ah, even
1175 passing the list together with a list of indexes seems feasiable.
1177 Or maybe ask for the inverted list?
1179 Whenever the complete array is covered by the interval we say we are
1180 fully covered and if the recentfile is not expired, we are uptodate.
1182 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1184 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1186 * need a way to "return" the next entry after the end of a list. When
1187 the caller says "before" or "after" we would like to know if he could
1188 cover that interval/threshold or not because this influences the effect
1189 of a newer timestamp of that recentfile. DONE with $opt{info}.
1191 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1193 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1195 loop implies tixu (?). Nope, something like --statefile decides. Per
1196 default we do ...?
1198 T test, I init, X including removals, U nodirtymark
1200 So we have no concept of dirtymarks, we only trust that since we are
1201 running we have observed everything steadily. But people will not let
1202 this program run forever so we must consider both startup penalty and
1203 book keeping for later runs. We keep this for later. For now we write a
1204 long running mirror that merges several intervals.
1206 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1208 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1209 the intervals!
1211 * rersyncrecent, the script: default to one week. The name of the switch
1212 is --after. Other switches? --loop!
1214 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1216 * need a switch --skip-deletes (?)
1218 * need a switch --enduser that tells us that the whole tempfile
1219 discipline is not needed when there is no downstream user. (?)
1221 Without this switch we cannot have a reasonable recent.pl that just
1222 displays the recent additions. Either we accept to download everything.
1223 Or we download temporary files without the typical rsync protocol
1224 advantages.
1226 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1227 File::Temp::tempdir, this might be a win.
1229 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1231 * apropos missing: we have no push, we never know the downstream
1232 servers. People who know their downstream hosts and want to ascertain
1233 something will want additional methods we have never thought about, like
1234 update or delete a certain file.
1236 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1238 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1239 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1240 difference.
1242 * rsync options: it's a bit of a pain that we usually need several rsync
1243 options, like compress, links, times, checksum and that there is no
1244 reasonable default except the original rsync default. I think wee can
1245 safely assume that the rsync options are shared between all recentfile
1246 instances within one recent tree.
1248 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1250 * deletes: if a delete follows an add quickly enough it may happen that
1251 a downstream mirror did not see the add at all! It seems this needs to
1252 be mentioned somewhere. The point here is that even if the downstream is
1253 never missing the principal timeframe it may encounter a "delete" that
1254 has no complimentary "add" anywhere.
1256 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1258 * I suspect the treat of metadata is incorrect during read or something.
1259 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1260 contained more than 6 hours worth of data. At 06:08 we merged into the
1261 1d file. We need to take snapshots of the 6h file over the course of an
1262 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1263 enough.
1265 Much worse: watching the 1h file: right at the moment (at 06:35) it
1266 covers 1218867584-1219120397 which is 70 hours.
1268 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1269 08:19:44 2008, that is when I checked out last time, so it seems to be
1270 aggregating and never truncating?
1272 No, correct is: it is never truncating; but wrong is: it is aggregating.
1273 It does receive a lot of events from time to time from a larger file.
1274 Somehow a large file gets merged into the small one and because the
1275 "meta/merged" attribute is missing, nobody is paying attention. I
1276 believe that I can fix this by making sure that metadata are honoured
1277 during read. DONE and test adjusted.
1279 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1281 * grand renaming plan
1283 remotebase => remoteroot to fit well with localroot DONE
1284 local_path() => localroot seems to me should already work DONE
1285 recentfile_basename => rfilename no need to stress it has no slash DONE
1287 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1288 kern? It would anyway need a deprecation cycle because it is an
1289 important constructor.
1291 * I like the portability that Data::Serializer brings us but the price
1292 is that some day we might find out that it is slowing us a bit. We'll
1293 see.
1295 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1297 * should we not enter the interval of the principal (or the interval of
1298 the merging file?) in every aggregated/merged file?
1300 * we should aim at a first release and give up on thinking about
1301 sanitizing stuff and zloop. Let's just admit that a full traditional
1302 rsync is the only available sanitizer ATM. Otherwise it's complicated
1303 stuff: sanitizing on the origin server, sanitizing on the slaves,
1304 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1305 get the basics out before this becomes a major cause for mess.
1307 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1309 * On OSes not supporting symlinks we expect that RECENT.recent contains
1310 the contents of the principal recentfile. Actually this is identical on
1311 systems supporting symlinks. Simple, what follows from that is that we
1312 need to keep the serializer in the metadata because we cannot read it
1313 from the filename, doesn't it? Of course not. It's a chicken and egg
1314 problem. This leaves us with the problem to actually parse the
1315 serialized data to find out in which format it is. So who can do the 4
1316 or 5 magics we wanted to support? File::LibMagic?
1318 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1320 * remotebase and recentfile_basename are ugly names. Now that we need a
1321 word for the shortest/principal/driving recentfile too we should do
1322 something about it.
1324 localroot is good. rfile is good. local_path() is bad, local_path($path)
1325 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1326 already deprecated.
1328 Up to now remotebase was the string that described the remote root
1329 directory in rsync notation, like pause.perl.org::authors. And
1330 recentfile_basename was "RECENT-1h.yaml".
1332 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1334 * The test that was added in today's checkin is a good start for a test
1335 of rmirror. We should have more methods in Recent.pm: verify,
1336 addmissingfiles. We should verify the current tree, then rmirror it and
1337 then verifytree the copy. We could then add some arbitrary file and let
1338 it be discovered by addmissingfiles, then rmirror again and then
1339 verifytree the copy again.
1341 Then we could start stealing from csync2 sqlite database [no port to
1342 OSX!] and fill a local DB. And methods to compare the database with the
1343 recentfiles. Our strength is that in principle we could maintain state
1344 with a single float. We have synced up to 1234567890.123456. If the Z
1345 file does not add new files all we have to do is mirror the new ones and
1346 delete the goners.
1348 This makes it clear that we should extend current protocol and declare
1349 that we cheat when we add files too late, just to help the other end
1350 keeping track. Ah yes, that's what was meant when zloop was mentioned
1351 earlier.
1353 Maybe need to revisit File::Mirror to help me with this task.
1355 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1357 * There must be an allow-me-to-truncate flag in every recentfile.
1358 Without it one could construct a sequence of updates winning the locking
1359 battle against the aggregator. Only if an aggregator has managed to
1360 merge data over to the next level, truncating can be allowed. DONE with
1361 accessor merged.
1363 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1365 * We should probably guarantee that no duplicates enter the aggregator
1366 array.
1368 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1370 * To get merge operation faster would need a good benchmark test. What
1371 02 spits out isn't reliable enough and is dominated by many other
1372 things. Between
1374 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1375 Author: Andreas J. Koenig <k@k75.(none)>
1376 Date: Sat Aug 2 07:58:04 2008 +0200
1378 and
1380 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1381 Author: Andreas J. Koenig <k@k75.(none)>
1382 Date: Sat Aug 2 11:40:29 2008 +0200
1384 there was a lot of trying but the effect is hardly measurable with
1385 current tests.
1387 * overhead of connecting seems high. When setting
1388 max_files_per_connection to 1 we see that.
1390 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1392 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1394 25138.2215600014/3600 = 6.98283932222261
1396 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1397 be a bug in the aggregator. FIXED
1399 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1401 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1402 we have it, can we mirror it, mirror it!
1404 I fear this needs a new class which might be called
1405 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1406 and treat them as an entity. I realize that a single recentfile may be
1407 sufficient for certain tasks and that it is handy for the low level
1408 programmer but it is not nice to use. If there is a delete in the 1h
1409 file then the 6h file still contains it. Seekers of the best information
1410 need to combine at least some of the recentfiles most of the time.
1412 There is the place for the Z loop!
1414 But the combination is something to collect in a database, isn't it. Did
1415 csync2 just harrumph?
1417 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1419 * it just occurred to me that hosts in the same mirroring pool could
1420 help out each other even without rewriting the recentfile. Just fetch
1421 the stuff to mirror from several places, bingo. But that's something
1422 that should rather live in a separate package or in rsync directly.
1424 * cronjobs are unsuited because with ntp they would all come at the full
1425 minute and disturb each other. Besides that I'd hate to have a backbone
1426 with more than a few seconds latency.
1428 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1430 * a second rsync server with access control for PAUSE. Port? 873 is the
1431 standard port, let's take 8873.
1433 * if there were a filesystem based on this, it would have a slow access
1434 to inexistent files. It would probably provide wrong readdir (only based
1435 on current content) or also a slow one (based on a recentfile written
1436 after the call). But it would provide fast access to existing files. Or
1437 one would deliberately allow slightly blurred answers based on some
1438 sqlite reflection of the recentfiles.
1440 * todo: write a variant of mirror() that combines two or more
1441 recentfiles and treats them like one
1443 * todo: signal handler to remove the tempfile
1445 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1447 * now that we have the symlink I forgot how it should be used in
1448 practice.
1450 * the z loop: add missing files to Z file. Just append them (instead of
1451 prepending). So one guy prepends something from the Y file from time to
1452 time and another guy appends something rather frequently. Collecting
1453 pond. When Y merges into Z, things get epoch and the collecting pond
1454 gets smaller. What exactly are "missing files"?
1456 take note of current epoch of the alpha file, let's call it the
1457 recent-ts
1459 find all files on disk
1461 remove all files registered in the recentworld up to recent-ts
1463 remove all files that have been deleted after recent-ts according to
1464 recentworld
1466 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1468 * rersyncrecent might be a cronjob with a (locked) state file which
1469 contains things like after and maybe last z sync or such?
1471 rrr-mirror might be an alternative name but how would we justify the
1472 three Rs when there is no Re-Rsync-Recent?
1474 With the --loop parameter it is an endless loop, without it is no loop.
1475 At least this is simple.
1477 * todo: new accssor z-interval specifies how often the Z file is updated
1478 against the filesystem. We probably want no epoch stamp on these
1479 entries. And we want to be able to filter the entries (e.g. no
1480 by-modules and by-category tree)
1482 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1484 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1485 into Z. We do not need the other files filled up with filesystem
1486 contents though.
1488 * need interface to query for a file in order to NOT call update on
1489 PAUSE a second time within a short time.
1491 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1493 * recommended update interval? Makes no sense, is different for
1494 different users.
1496 * Moosify
1498 Local Variables:
1499 mode: change-log
1500 change-log-default-name: "Todo"
1501 tab-width: 2
1502 left-margin: 2
1503 End: