| blob | 8f0b14f093746a39cca4349f3c8bcda4da5f059a |
1 2011-02-17 Andreas J. Koenig <andk@cpan.org>
3 * .recent/ or .rrr/ for metadata
5 * ALERT rrr: we have a piece of code somewhere in the mirror()
6 subroutine that reads the raw YAML file with normal open() and cuts it
7 off for efficiency, then feeds it to yaml::load in order to get some
8 metadata. Reading the thing with the whole tail of the recent array
9 kills the whole idea. The whole thing will not work with JSON.
11 So we must invent a protocol 3 that works with two index files.
13 2011-02-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
15 * I think interval guaranteed with the $last_aggregate_call
16 variable (both in fsck and server) need to be fixed with a $recc
17 variable to something like 60 seconds or even less.
19 * minor bug: fsck just added the lockfile to the index which should be
20 considered bookkeeping.
22 * speed: a large directory remove that is done by the kernel in 1
23 seconds triggers minutes of bookkeeping (07:56:42 - 08:01:10 for ~3000
24 files, all due to Schlemiehl again. Want to do some collecting of
25 pending ops before actually writing to the RECENT files. Lock!
27 Repeating the timing with 2600 files and it took 21:28:08 - 21:35:27.
29 Now having rewritten the loop to use batch_update(): 4383 file removed
30 21:43:37 - 21:44:12. From 7:30 to 0:35 while doing 60% more work that's
31 a joy.
33 * lockdirectory expiration? server died and blocked fsck for so long.
35 * At the moment rrr-init is not needed anymore, rrr-server can be
36 started and followed by rrr-fsck. Still stupid but not the main
37 showstopper.
39 2011-02-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
41 * IN_Q_OVERFLOW needs to trigger an fsck, likewise the entry into the
42 server.
44 2011-01-30 Andreas J. Koenig <andk@cpan.org>
46 * rrr-init and rrr-server need to be used such:
48 init
49 server
50 init again
52 Stupid but no showstopper ATM.
54 2010-10-26 Andreas J. Koenig <andk@cpan.org>
56 * forking slave is done, needs more testing.
58 2010-07-06 Andreas J. Koenig <andk@cpan.org>
60 * The two next most important goals are to get a useable rrr-server and
61 a forking slave implementation that uses virtually no memory after the
62 first pass through. Does it need the status file?
64 * rrr-init now being fast we can now remove the alpha alerts. Done.
66 2010-06-16 Andreas J. Koenig <andk@cpan.org>
68 * rrr-init seems to be sssslllloooowwww. At 3:49 I started it for the
69 video tree. After twelve minutes the file has reached a size of 500k. It
70 is locking and rewriting the YAML after every file. Wasn't there an
71 unlocked variant? For "init" one would not expect locked correctness and
72 a rewrite after every item. Given that this tends to become slower with
73 every file it is really unacceptable.
75 Reminds me of rrr-dirtyupdate (the former name of rrr-update). It seems
76 the code was not dirty enough and that was the reason why it was so
77 slow. And rrr-update does not have a --dirty switch, so it looks like we
78 have no way to force dirty and fast operation yet but we need it for
79 init.
81 Irritating the bootstrapping race: the order {(1) fill initial
82 array, (2) start server} leaves the need for an fsck to cover the time
83 between 1 and 2. Given that regular fsck is needed in any case the price
84 is not high but still doesn't give a warm feeling. Maybe starting the
85 server will by itself do an fsck?
87 2010-06-15 Andreas J. Koenig <andk@cpan.org>
89 * remove the wild alpha alerts and say, it is alpha but parts of it are
90 already working very well or so.
92 2009-12-10 Andreas J. Koenig <andk@cpan.org>
94 * Prominent request for statusfile to safe memory consumption. I'd like
95 to work backwards: get a verification program that verifies how well in
96 sync two nodes are without using frmr itself. Pretend that the forking
97 agent already works and can be interrupted with ^C. Then let it die at
98 random and restart at random and see the memory consumption and the
99 failure rate over time. Compare it with the curve we have now.
101 Of course we also invented the status file as a sharp debugging tool.
102 I'd like to see it in action. See also: F:R:M:Recent::thaw().
104 2009-06-22 Andreas J. Koenig <andk@cpan.org>
106 * Todo: rrr-init to generate some index files so that rrr-server can be
107 started.
109 2009-05-02 Andreas J. Koenig <andk@cpan.org>
111 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
112 so we can use it for normal update with a single file. And add a cronjob
113 that runs rrr-fsck so we get a reminder when something is not working
114 as it should.
116 2009-04-27 Andreas J. Koenig <andk@cpan.org>
118 * Todo: I still see more link_stat errors than expected. I come back to
119 the former suspicion that we do not do necessary bookkeeping on deleted
120 files so that when we reach the "new" event on a meanwhile deleted file
121 we try to mirror it although we could know better. It's just noise, no
122 harm but noise is irritating.
124 * Todo: more experimenting with runstatusfile to get it official
126 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
127 parse the error.
129 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
131 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
133 And that ignore_link_stat_error is propagated in sparse_clone.
135 DONE
137 For the record: the bug was that ignore_link_stat_error was lost during
138 sparse_clone. And then the second bug was that it did not default to
139 true. Both are now fixed.
141 2009-04-25 Andreas J. Koenig <andk@cpan.org>
143 * Todo: fill rrr program with life
145 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
147 * Minor bug: I see complaints about files not existing like
148 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
149 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
151 In the RECENT files these are delete events but of course also new
152 events in older recent files.
154 I think we do not properly remember deletes when we skip-deletes and so
155 do not filter them out when we later see the "new" event.
157 Yes, but that what ignore_link_stat_error is for, we know about this
158 race condition. We must make ignore_link_stat_error default to true.
160 DONE
162 * bug in Done.pm failing to merge two fields into one when a third is
163 present. FIXED and accompanied with new test.
165 2009-04-24 Andreas J. Koenig <andk@cpan.org>
167 * Todo: keep the index clean and run some kind of nondestructive fsck 4
168 times a day.
170 * who is our backbone?
172 http://cpan.cpantesters.org/authors/02STAMP # barbie
173 http://cpan.solfo.com/authors/02STAMP # abh
174 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
176 2009-04-17 Andreas J. Koenig <andk@cpan.org>
178 * the changelog helper I included in the Makefile release memo does not
179 show tags and is much less useful than I thought. gitk is probably much
180 more convenient.
182 2009-04-16 Andreas J. Koenig <andk@cpan.org>
184 * ABH writes:
186 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
187 #### Ydr_.yaml ... DONE
188 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
189 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
190 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
191 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
192 #### gone already at cpan-pause.pl line 0
194 Running without skip-deletes now but cannot reproduce.
196 2009-04-15 Andreas J. Koenig <andk@cpan.org>
198 * consider state directory such we can restart after a ^C or come again
199 from a cronjob.
201 Reuse _runstatusfile: write atomically, see how we can use the dump to
202 start running again. Do not leak and remove globs before writing. Do not
203 only dump one rf, dump the whole complex.
205 * consider the effect of ^C. Is it safe?
207 * want to use Perl::Version to up the versions on every release. Only if
208 things have changed, of course. Does this actually work?
210 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
211 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
213 Must do the bookkeeping myself when this should happen, maybe based on git?
215 I think I prefer the same version number for all pm files. But one also
216 needs a routine for setting them. Ahh:
218 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
220 NEEDSMOREWORK
222 * why indexer wrong?
224 cpan[2]> m /Mirror::Recent/
225 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
226 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
227 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
228 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
229 4 items found
231 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
233 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
235 I just tried MM 6.5102 and it again produces only a provides for Recent,
236 not for the other modules.
238 My own fault. FIXED in the Makefile.PL.
240 2009-04-13 Andreas J. Koenig <andk@cpan.org>
242 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
244 * is it true (as stated at
245 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
246 that rsync can lead to errors when upstream changes before client sync
247 has completed? Can we see the error when we run this on upstream:
249 % perl -e '
250 use Time::HiRes qw(time);
251 while (){
252 open my $fh, ">", "changingfile.txt.new" or die;
253 print $fh (time."\n") x 1000000;
254 close $fh;
255 rename "changingfile.txt.new", "changingfile.txt" or die;
256 }
257 '
259 And on the receiving end:
261 % while true; do
262 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
263 sleep 1;
264 done
266 I cannot get it to fail. Apparently it is sufficient when upstream
267 always writes atomically (which of course is mandatory)?
269 SENT to mailing list Apr 28.
271 * from ABH:
273 > https://fedorahosted.org/InstantMirror/
274 > https://www.redhat.com/mailman/listinfo/instantmirror-list
275 > irc.freenode.net channel #instantmirror
277 A cool name for a project. Inspires me to write a few sentences about
278 bittorrent's role in the grand picture.
280 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
281 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
283 2009-04-12 Andreas J. Koenig <andk@cpan.org>
285 * Interesting last minute bug during real download testing: the output
286 isn't as pretty anymore, it seems that more work is being done than
287 needed.
289 Introducing a highlevel _alluptodate method helped a lot in debugging
290 this.
292 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
293 different on 1h file than on the other files. This lead to frequent
294 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
295 issue.
297 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
299 * Todo: try if we can be faster with a native float package. I'm really
300 glad to have a machine exposing perl floating point bugs, it forced me
301 to retract this item from the changes file: "several speedups in the
302 fakebigfloat code" before I get cpan testers fails.
304 2009-04-11 Andreas J. Koenig <andk@cpan.org>
306 * Bug: when the dirtymark changes upstream, then the downstream server
307 notices it immediately and mirrors the "1h" file but then leaves the
308 rmirror loop and continues with "6h" on the next round through. This
309 behaviour goes on until it reaches the Z file. Because we eat the Z file
310 piecemeal we leave the loop after a while but after that everybody seems
311 to have forgotten that there is some work left to be done. Bad, bad,
312 bad.
314 ETOOCONFUSING better name?
316 Recentfile::get_remote_recentfile_as_tempfile OK
317 Recentfile::resolve_recentfilename split_rfilename DONE
319 Recent::_principal_recentfile_object _principal_recentfile_fromremote
320 Recent::principal_recentfile OK
321 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
323 some void, some not, some not void but called in void context.
325 2009-04-10 Andreas J. Koenig <andk@cpan.org>
327 * Bug? When the dirtymark changes then the second tier hosts quickly
328 reset their DONE state and restart mirroring. But then they mirror
329 potentially outdated index files with inconsistent dirtymark because the
330 first tier box may be a bit behind on the large recentfiles.
332 This means we need a brake when iterating over the recentfiles that
333 refuses to use a recentfile with the wrong dirtymark.
335 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
337 * install the bin/* files and talk about rrr-overview somewhere.
339 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
340 created outside the target tree.
342 * Barbie asks for better logging capabilities
344 2009-03-29 Andreas J. Koenig <andk@cpan.org>
346 * known bug: cannot configure the Recentfile objects to keep delete
347 events. How to call the accessor for that? keep_delete_objects_forever?
349 Sounds acceptable. FIXED (but untested).
351 * known bug: keeps temporary index files lying around.
353 2009-03-24 Andreas J. Koenig <andk@cpan.org>
355 * Some equivalent for
357 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
359 ? DONE
361 2009-03-22 Andreas J. Koenig <andk@cpan.org>
363 * k81 is again client of k75 with different parameters.
365 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
366 about it all the time and nobody reacts accordingly; (2) lots of temp
367 files get created and not removed; culprit the new call to
368 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
369 caller has to remove the tempfile after use.
371 Need the drawing board.
373 PARTLY FIXED: retracting the idea to call
374 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
375 and unseed such that uptodate now means "we have mirrored this
376 recentfile's bunch of files" and seeded means "somebody believes the
377 index file for this rf needs be refreshed". This seems to work now.
379 Still collecting temp files that nobody cares to depollute.
381 2009-03-21 Andreas J. Koenig <andk@cpan.org>
383 * Bug?: should it be harder than it is atm to set the timestamp to the
384 future? YES, FIXED
386 * bug with native integers:
388 -
389 epoch: 997872011
390 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
391 type: new
392 -
393 epoch: 1195248431
394 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
395 type: new
397 Native integer broke when native math was turned off. FIXED
399 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
400 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
401 in this area.
403 Yes, records out of order:
405 447003 -
406 447004 epoch: 995885533
407 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
408 447006 type: new
409 447007 -
410 447008 epoch: 995890358
411 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
412 447010 type: new
413 447011 -
414 447012 epoch: 995892221
415 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
416 447014 type: new
418 FIXED with sanity check and later with the integer fix.
420 * Bug: want the index files in a .recent directory
422 * Bug: lots of dot files are not deleted in time
424 * possible test case: can a delete change the timestamp? This would
425 probably break the order of events.
427 2009-03-20 Andreas J. Koenig <andk@cpan.org>
429 * 1233701831.34486 what's so special about this number/string? It
430 belongs to
431 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
432 atm lives in Y,Q, and Z.
434 It is the first entry after id/--skip-locking which has timestamp
435 1234164228.11325 which represents a file that doesn't exist anymore.
437 * another bug:
439 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
440 _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
441 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
442 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
443 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
444 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
445 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
446 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
447 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
448 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
449 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
450 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
451 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
453 FIXED, it was the "--skip-locking" file where manual intervention was participating
455 * bug on the mirroring slave: when the dirtymark gets increased we
456 probably do not reset the done intervals. The mirrorer stays within
457 tight bounds where it tries to sync with upstream and never seems to
458 finish. In the debugging state file I see lots of identical intervals
459 that do not get collapsed. When I restart the mirrorer it dies with:
461 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
462 _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
463 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
464 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
465 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
466 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
467 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
468 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
469 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
470 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
472 and debugging stands at
474 At: before
475 Brfinterval: Z
476 Ci: 227
477 Dre+1:
478 epoch: 1237400802.5789
479 path: id/J/JO/JOHND/CHECKSUMS
480 type: new
481 Dre-0:
482 epoch: 1237400807.97514
483 path: id/M/MI/MIYAGAWA/CHECKSUMS
484 type: new
485 Dre-1:
486 epoch: 1237400817.94363
487 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
488 type: new
489 Eintervals:
490 -
491 - 900644040
492 - 900644040
493 - []
495 and it is reproducable.
497 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
498 pause has a fresh one.
500 FIXED, it was the third anded term in each of the ifs in the IV block in
501 _register_one: with that we make sure that we do not stamp on valuable
502 interval data.
504 2009-03-17 Andreas J. Koenig <andk@cpan.org>
506 * done: verified the existence of the floating point bug in bleadperl
507 and verified that switching from YAML::Syck to YAML::XS does not resolve
508 it.
510 BTW, the switch was doable with
512 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
514 and should be considered as a separate TODO
516 * todo: integrate a dirty update with two aggregate calls before
517 unlocking for frictionless dirtying DONE
519 * todo: start the second rsync daemon on pause DONE
521 * todo: move index files to .recent: this cannot simply be done by
522 setting filenameroot to .recent/RECENT. Other parts of the modules rely
523 on the fact that dirname(recentfile) is the root of the mirrored tree.
525 2009-03-16 Andreas J. Koenig <andk@cpan.org>
527 * What was the resolution of the mirror.pl delete hook bug? Do we call
528 the delete hook when pause removes a file from MUIR?
530 * Today on pause: Updating 2a13fba..29f284d and installing it for
531 /usr/local/perl-5.10.0{,-RC2}
533 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
534 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
536 2009-03-15 Andreas J. Koenig <andk@cpan.org>
538 * currently recent_events has the side effect of setting dirtymark
539 because it forces a read on the file. That should be transparent, so
540 that the dirtymark call always forces a cache-able(?) read.
542 * The bug below is -- after a lot of trying -- not reproducible on a
543 small script, only in the large test script. The closest to the output
544 below was:
546 #!perl
547 use strict;
548 use Devel::Peek;
549 my $x = "01237123229.8814";
550 my($l,$r);
551 for ($l,$r) {
552 $_ = "x"x34;
553 }
554 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
555 $r = int $r;
556 $l = "1237123231.22458";
557 $r = "1237123231.22458";
558 1 if $l/1.1;
559 Devel::Peek::Dump $l;
560 Devel::Peek::Dump $r;
561 Devel::Peek::Dump $x = $l <=> $r;
562 die "BROKE" if $x;
563 __END__
565 The checked in state at c404a85 fails the test with my
566 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
567 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
568 in the test script always consistent.
570 * Todo: write a test that inserts a second dirty file with an already
571 existing timestamp. DONE
573 * Bug in perl 5.10 on my 64bit box:
575 DB<98> Devel::Peek::Dump $l
576 SV = PVMG(0x19e0450) at 0x142a550
577 REFCNT = 2
578 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
579 IV = 0
580 NV = 1237123231.22458
581 PV = 0x194ce70 "1237123231.22458"\0
582 CUR = 16
583 LEN = 40
585 DB<99> Devel::Peek::Dump $r
586 SV = PVMG(0x19e0240) at 0x142a3e8
587 REFCNT = 2
588 FLAGS = (PADMY,POK,pPOK)
589 IV = 1237123229
590 NV = 1237123229.8814
591 PV = 0x19ff900 "1237123231.22458"\0
592 CUR = 16
593 LEN = 40
595 DB<100> Devel::Peek::Dump $l <=> $r
596 SV = IV(0x19ea6e8) at 0x19ea6f0
597 REFCNT = 1
598 FLAGS = (PADTMP,IOK,pIOK)
599 IV = -1
601 DB<101> Devel::Peek::Dump $l
602 SV = PVMG(0x19e0450) at 0x142a550
603 REFCNT = 2
604 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
605 IV = 0
606 NV = 1237123231.22458
607 PV = 0x194ce70 "1237123231.22458"\0
608 CUR = 16
609 LEN = 40
611 DB<102> Devel::Peek::Dump $r
612 SV = PVMG(0x19e0240) at 0x142a3e8
613 REFCNT = 2
614 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
615 IV = 1237123231
616 NV = 1237123231.22458
617 PV = 0x19ff900 "1237123231.22458"\0
618 CUR = 16
619 LEN = 40
621 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
623 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
625 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
627 * $obj->merge ($other) needs to learn about equal epoch which may happen
628 since dirty_epoch intruded.
630 * Wontfix anytime soon: I think we currently do not support mkdir. Only
631 files!
633 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
635 * Todo: continue working on update(...,$dirty_epoch). It must be
636 followed by a fast_aggregate! DONE
638 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
640 * maybe we need a closest_entry or fitting_interval or something like
641 that. We want to merge an event into the middle of some recentfile.
642 First we do not know which file, then we do not know where to lock,
643 where to enter the new item, when and where to correct the dirtymark.
645 So my thought is we should first find which file.
647 Another part of my brain answers: what would happen if we would enter
648 the new file into the smallest file just like an ordinary new event,
649 just as an old event?
651 (1) we would write a duplicate timestamp? No, this would be easy to
652 avoid
654 (2) we would make the file large quickly? Yes, but so what? We are
655 changing the dirtymark, so are willing to disturb the downstream hosts.
657 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
659 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
661 leak!
663 https://rt.cpan.org/Ticket/Display.html?id=41199
665 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
667 Strangest output being HAKANARDO who managed to upload
669 Here is a better oneliner that includes also the first line of each
670 finding:
672 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
674 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
676 * memory leak in the syncher? It currently weighs 100M.
678 Update 2008-11-02:
680 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
683 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
685 * lookup by epoch and by path and use this ability on the pause to never
686 again register a file twice that doesn't need it. Let's call it
687 contains().
689 * after the dirtymark is done: fill up recentfiles with fake (historic)
690 entries; fill up with individual corrections; algorithm maybe to be done
691 with bigfloat so that we can always place something in the middle
692 between two entries. Before we must switch to bigfloat we could try to
693 use Data::Float::nextup to get the.
695 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
696 directly.
698 * dirtymark now lives in Recentfile, needs to be used in rmirror.
700 * find out why the downloader died after a couple of hours without a net
701 connection. Write a test that survives the not-existence of the other
702 end forever.
704 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
706 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
707 it makes less sense with tight coupling for secondary files. But asking
708 the server what the current epoch is might be cheaper on HTTP than on
709 rsync. (Needs to be evaluated)
711 * remove the 0.00 from the verbose overview in the Merged column in the
712 Z row. DONE
714 * write tests that expose the problems of the last few days: cascading
715 client/server roles, tight coupling for secondary RFs, deletes after
716 copies.
718 * Some day we might want to have policy options for the slave:
719 tight/loose/no coupling with upstream for secondary RFs. tight is what
720 we have now. loose would wait until a gap occurs that can be closed.
722 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
724 * revisit all $rfs->[$i+1] places if they now make sense still
726 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
728 * another bug is the fact that the mirror command deletes files before
729 it unhides the index file, thus confusing downstream slaves. We must not
730 delete before unhiding and must delete after unhiding. FIXED.
732 * new complication about the slave that is playing a server role.
733 Currently we mirror from newest to oldest with a hidden temporary file
734 as index. And when one file is finished, we unhide the index file.
735 Imagine the cascading server/slave is dead for a day. It then starts
736 mirroring again with the freshest thing and unhides the freshest index
737 file when it has worked through it. In that moment it exposes a time
738 hole. Because it now works on the second recentfile which is still
739 hidden.
741 We currently do nothing special to converge after such a drop out. At
742 least not intentionally and robustly and thought through.
744 The algorithm we use to seed the next file needs quite a lot of more
745 robustness than it currently has. Something to do with looking at the
746 merged element of the next rf and when it has dropped off, we seed
747 immediately. And if it ramains dropped off, we seed again, of course.
749 Nope, looking from smaller to larger RFS we look at the merged element
750 of this RF and at the minmax/max element of the next RF. If that
751 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
752 seeding it.
754 And we need a public accessor seed and unseed or seeded. But not the mix
755 of public and private stuff that then is used behind the back.
757 And then the secondary* stuff must go.
759 And we must understand what the impact is on the DONE system. Can it go
760 unnoticed that there was a hole? And could the DONE system have decided
761 the hole is covered? This should be testable with three directories where
762 the middle stops working for a while. Done->merge is suspicious, we must
763 stop it from merging non-conflatable neighbors due to broken continuity.
765 FIXED
767 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
769 * Slaven suggests to have the current epoch or the whole current
770 recentfile available from the HTTP server and take it away with
771 keepalive. This direction goes the granularity down to subseconds.
773 We might want to rewrite everything to factor out transport and allow
774 the whole thing to run via HTTP.
776 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
778 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
779 symlink?
781 * smoker on k81 fetching from k75 to verify cascading works. See
782 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
784 * maybe the loop should wait for CHECKSUMS file after every upload. And
785 CPAN.pm needs to deal with timestamps in the future.
787 * do not forget the dirtymark!
789 Text: have a new flag on recentfiles with the meaning: if this
790 changes, you're required to run a full rsync over all the files. The
791 reason why we set it would probably be: some foul happened. we injected
792 files in arbitrary places or didn't inject them although they changed.
793 The content of the flag? Timestamp? The relation between the
794 recentfiles would have to be inheritance from the principal, because any
795 out of band changes would soon later propagate to the next recentfile.
797 By upping the flag often one can easily ruin the slaves.
799 last out of band change? dirtymark?
801 Anyway, this implies that we read a potentially existing recentfile
802 before we write one.
804 And it implies that we have an eventloop that keeps us busy in 2-3
805 cycles, one for current stuff (tight loop) and one for the recentfiles
806 (cascade when principal has changed), one for the old stuff after a
807 dirtymark change.
809 And it implies that the out-of-band change in any of the recentfiles
810 must have a lock on the principal file and there is the place to set the
811 dirtymark.
813 * start a FAQ, especially quick start guide questions. Also to aid those
814 problematic areas where we have no good solution, like the "links"
815 option to rsync.
817 * wish feedback when we are slow.
819 * reduce mccabe
821 * Remove a few DEBUG statements.
823 * The multiple-rrr way of doing things needs a new option to rmirror,
824 like piecemeal or so. Not urgent because after the first pass through,
825 things run smoothely. It's only ugly during the first pass.
827 * I have the suspicion that the code is broken that decides if the
828 neighboring RF needs to be seeded. I fear when too much time has gone
829 between two calls (in our case more than one hour), it would not seed
830 the neighbor. Of course this will never be noticed, so we need a good
831 test for it.
833 * local/localroot confusion: I currently pass both options but one must
834 do.
836 * accounts for early birds on PAUSE rsync daemon.
838 * hardcoded 20 seconds
840 * who mirrors the index? DOING now.
842 * which CPAN mirrors offer rsync?
844 * visit all XXX, visit all _float places
846 * rename the pathdb stuff, it's too confusing. No idea how.
848 * rrr-inotify, backpan, rrr-register
850 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
852 * current bugs: the pathdb seems to get no reset, the seeding of the
853 secondaryttl stuff seems not to have an effect. Have helped myself with
854 a rand(10), need to fix this back. So not checked in. Does the rand
855 thing even help?
857 The rand thing helps. The secondaryttl stuff was in the wrong line,
858 fixed now.
860 The pathdb stuff was because I called either _pathdb or __pathdb on the
861 wrong object. FIXED now.
863 * It's not so beautiful if we never fetch the recentfiles that are not
864 the principal, even if this is correct behaviour. We really do not need
865 them after we have fetched the whole content.
867 OK, we want a switch for that: secondaryttl DONE
869 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
871 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
872 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
873 it lets 40 follow. FIXED
875 * See that the long running process really only updates the principal
876 file unless it has missed a timespan during which something happened. If
877 nothing happened, it must notice even when it misses the timespan. DONE
879 * we must throw away the pathdb when we have reached the end of Z. From
880 that moment we can have a very small pathdb because the only reason for
881 a pathdb is that we know to ignore old records in old files. We won't
882 need this pathdb again before the next full pass over the data is
883 necessary and then we will rebuild it as we go along. DONE
885 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
887 * I think, Done::register_one is doing wrong in that it does not
888 conflate neighboring pieces. The covered() method cannot do this because
889 it has no recent_events array at hand. But register_one has it and could
890 do it and for some reason misses to do it (sometimes).
892 This means that the three tests I just wrote can probably not survive
893 because they test with an already broken Done structure.
895 The art now is to detect how it happens, then to reproduce, then write a
896 test, then fix it.
898 So from the logfile this is what happens: we have a good interval with
899 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
900 top of it. Locally we now mirror F2 and open a new done interval for it.
901 Then we mirror F1 but this time with the timestamp T1b. And when we then
902 try to close the gap, we do not find T1 but instead something older. We
903 should gladly accept this older piece and this would fix this bug.
905 FIXED
907 * bug to fix: when the 1h file changes while rmirror is running, we do
908 correctly sync the new files but never switch to the 6h file but rather
909 stay in a rather quick loop that fetches the 1h file again and again.
911 Is it possible that we initialize a new object? Or does
912 get_remote_recentfile_as_tempfile overwrite something in myself?
914 Want a new option: _runstatusfile => $file which frequently dumps the
915 state of all recentfiles to a file.
917 FIXED
919 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
921 * Todo: now teach update to verify the timestamp is about to write
922 against the previous and use _increase_a_bit if it doesn't comply with
923 strict monotony. DONE
925 * The problem of rounding. So far perl's default precision was
926 sufficient. One day it won't be. FakeFloat has an easy job when it is
927 only reading and other machines have written correctly. But when we want
928 to write a floating point number that is a bit larger than the other
929 one, then we need our own idea of precision.
931 Slaven said: just append a "1". This might be going towards the end of
932 usability too quickly. I'd like something that actually uses the decimal
933 system. Well, appending a 1 also does this but...
935 E.g. we have 1.0. nextup on this architecture is starting with
936 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
937 taken the 1.0000000000000003 and the next user comes and the time tells
938 him 1.0 again. He has to beat my number without stepping over the
939 nextup. This is much less space than I had when I chose 1,2,3.
941 What is also irritating is that nextup is architecture dependent. The
942 128 bit guy must choose very long numbers to fit in between whereas the
943 other one with 16 bit uses larger steps. But then the algorithm is the
944 same for both, so that would be a nice thing.
946 I see two situation where we need this. One is when Time::HiRes returns
947 us a value that is <= the last entry in our recentfile. In this case
948 (let's call it the end-case) we must fill the region between that number
949 and the next higher native floating point number. The other is when we
950 inject an old file into an old recentfile (we would then also set a new
951 dirtymark). We find the integer value already taken and need a slightly
952 different one (let's call it the middle-case). The difference between
953 the two situations is that the next user will want to find something
954 higher than my number in the end-case and something lower than my number
955 in the middle case.
957 So I suggest we give the function both a value and an upper bound and it
958 calculates us a primitive middle. The upper bound in the middle-case is
959 the next integer. The upper bound on the end-case is the nextup floating
960 point number. But the latter poses another problem: if we have occupied
961 the middle m between x and nextup(x), then the nextup(m) will probably
962 not be the same as nextup(x) because some rounding will take place
963 before the nextup is calculated and when the rounding reaches the
964 nextup(x), we will end up at nextup(nextup(x)).
966 So we really need to consider the nextup and the nextdown from there and
967 then the middle and that's the number we may approach asymptotically.
968 Ugly. But DONE.
970 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
972 * consider deprecating the use of RECENT.recent as a symlink. It turns
973 out to need extra hoops with the rsync options and just isn't worth it.
974 Or maybe these extra hoops are needed anyway for the rest of the tree?
975 Nope, can't be the case because not all filesystems support symlinks.
977 But before doing the large step, I'll deprecate the call of
978 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
979 only introduced to resolve RECENT.recent and complicates the routine far
980 beyond what it deserves.
982 DONE. Won't deprecate RECENT.recent, just moved its handling to the
983 supervisor.
985 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
987 * I think it's a bug that the rsync_option links must be set to true in
988 order to support RECENT.recent and that nobody cares to set it
989 automatically. Similar for ignore_link_stat_errors.
991 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
993 * Todo: find all todos together and make a plan what is missing for a
994 release.
996 - verifytree or something like that. fsck maybe.
998 - rersyncrecent, the script itself? What it do?
1000 - a way to only mirror the recentfiles without mirroring the whole
1001 remote system such that people can decide to mirror only partially see
1002 also 2008-08-30. .shadow-xxx directory? this also needed for a
1003 filesystem that is still incomplete and might need the mirrorfiles for
1004 lookup(?)
1006 - long living objects that mirror again and again. Inject something
1007 into ta, see how it goes over to tb.
1009 - how do we continue filling up the DONE system when we use an object
1010 for the second time? "fully covered" and "uptodate" or new terminology.
1012 - overview called on the wrong file should be understandable
1014 - the meta data field that must change when we fake something up so that
1015 the downstream people know they have to re-fetch everything.
1017 - how tolerant are we against missing files upstream? how do we keep
1018 track? there are legitimate cases where we did read upstream index right
1019 before a file got deleted there and then find that file as new and want
1020 it. There are other cases that are not self healing and must be tracked
1021 and bugreported.
1023 - how, exactly, do we have to deal with deletes? With rsync errors?
1025 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
1026 authors) failed: No such file or directory (2)
1028 The file above is a delete in 1h and a new in file 1M and the
1029 delete in the locally running rmirror did not get propagated to the 1M
1030 object. Bug. And the consequence is a standstill.
1032 It seems that a slave that works with a file below the principal needs
1033 to merge things all the way up to get rid of later deletes. Or keep
1034 track of all deletes and skip them later. So we need a trackdeletes.pm
1035 similar to the done.pm?
1037 see also 2008-08-20 about spurious deletes that really have no add
1038 counterpart and yet they are not wrong.
1040 - consider the effect when resyncing the recentfile takes longer than
1041 the time per loop. Then we never rsync any file. We need to diagnose
1042 that and force an increase of that loop time. But when we later are fast
1043 enough again because the net has recovered, then we need to switch back
1044 to original parameters. ERm, no, it's enough to keep syncing at least
1045 one file before refetching an index file.
1047 - remember to verify that no temp files are left lying around and the
1048 signal handler
1050 - status file for not long running jobs that want to track upstream with
1051 a, say, cronjob.
1053 - revisit all XXX _float areas and study Sub::Exporter DONE
1055 - persistent DB even though we just said we do not need it. Just for
1056 extended capabilities and time savings when, for example, upstream
1057 announces a reset and we get new recentfiles and could then limit
1058 ourselves to a subset of files (those that have a changed epoch) in a
1059 first pass and would only then do the loop to verify the rest. Or
1060 something.
1062 * Todo: aggregate files should know their feed and finding the principal
1063 should be done stepwise. (?)
1065 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
1066 left locks around. DONE
1068 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1070 * maybe extend the _overview so that it always says if and where the
1071 last file is in the next file and where the next event in the next rf
1072 would lie. No, don't like this anymore. REJECT
1074 * take the two new redundant tests out again, only the third must
1075 survive. DONE
1077 * Todo: add a sanity check if the merged structure is really pointing to
1078 a different rf and that this different rf is larger. DONE
1080 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1082 * now test, if they are overlapping. And test if there is a file in the
1083 next rf that would fit into this rf's interval.
1085 1h 1222324012.8474 1222322541.7963 0.4086
1086 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
1087 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
1088 1W 1222313218.3626 1221708477.5829 167.9835
1090 I suspect that somebody writes a merged timestamp without having merged
1091 and then somebody else relies on it.
1093 If aggregate is running, the intervals must not be extravagated, if it
1094 is not running, there must not be bounds, the total number of events in
1095 the system must be counted and must be controlled throughout the tests.
1096 That the test required the additional update was probably nonsense,
1097 because aggregate can cut pieces too. FIXED & DONE
1099 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1101 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1102 changed. FIXED
1104 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1106 * Most apparent bug at the moment is that the recentfiles are fetched
1107 too often. Only the principal should be fetched and if it has not
1108 changed, the others should not be refetched. ATM I must admit that I'm
1109 happy that we refetch more often than needed because I can more easily
1110 fix bugs while the thing is running.
1112 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1113 done but the done system does not know about it. The reason for the
1114 failure is not known and we never reach the status uptodate because of
1115 this. We must get over it.
1117 Later it turns out that the origin server had a bug somewhere.
1118 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1119 array of the recentfile one position above. The bug was my own.
1121 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1123 * There is the race condition where the server does a delete and the
1124 slave does not yet know and then tries to download it because he sees
1125 the new. So for this time window we must be more tolerant against
1126 failure. If we cannot download a file, we should just skip it and should
1127 not retry immediately. The whole system should discover the lost thing
1128 later. Keeping track with the DONE system should really be a no brainer.
1130 But there is something more: the whole filesystem is a database and the
1131 recentfiles are one possible representation of it. It's a pretty useful
1132 representation I think that's why I have implemented something around
1133 it. But for strictly local operation it has little value. For local
1134 operation we would much rather have a database. So we would enter every
1135 recentfile reading and every rsync operation and for every file the last
1136 state change and what it leads to. Then we would always ignore older
1137 records without the efforts involved with recentfiles.
1139 The database would have: path,recentepoch,rsyncedon,deletedon
1141 Oh well, not yet clear where this leads to.
1143 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1145 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1146 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1148 * statusfile, maybe only on demand, alone to have a sharp debugging
1149 tool. It is locked and all recentfiles dump themselves into it and we
1150 can build a viewer that lets us know where we stand and what's inside.
1152 * remember: only the principal recentfile needs expiration, all others
1153 shall be expired by principal if it discovers that something has move
1154 upstream.
1156 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1158 * Always check if we stringify to a higher value than in the entry
1159 before. DONE
1161 * And in covered make an additional check if we would be able to see a
1162 numerical difference between the two numbers and if we can't then switch
1163 to a different, more expensive algorithm. Do not want to be caught by
1164 floating surprises. DONE
1166 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1168 * caching has several aspects here: we can cache the interval of the
1169 recentfile which only will change when the mtime of the file changes. We
1170 must re-mirror the recentfile when its ttl has expired. Does have_read
1171 tell you anything? It counts nothing at all. Only the mtime is
1172 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1173 thing the have_mirrored because it is unrelated to the mtime.
1175 * Robustness of floating point calculations! I always thought that the
1176 string calculated by the origin server for the floating representation
1177 of the epoch time is just a string. When we convert it to a number and
1178 later back to a string, the other computer might come to a different
1179 conclusion. This must not happen, we want to preserve it under any
1180 circumstances. I will have to write tests with overlong sequences that
1181 get lost in arithmetic and must see if all still works well. DONE
1183 But one fragile point remains: if one host considers a>b and the other
1184 one considers them == but no eq. To prevent this, we must probably do
1185 some extra homework. DONE
1187 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1189 * the concept of tracking DONE needs an object per recentfile that has
1190 something like these methods:
1192 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1194 covered() register() covered()
1196 The unclear thing is how we translate points in time into intervals. We
1197 could pass a reference to the current recent_events array when running
1198 we_have(xxx) and let the DONE object iterate over it such that it only
1199 has to store a list of intervals that can melt into each other. Ah, even
1200 passing the list together with a list of indexes seems feasiable.
1202 Or maybe ask for the inverted list?
1204 Whenever the complete array is covered by the interval we say we are
1205 fully covered and if the recentfile is not expired, we are uptodate.
1207 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1209 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1211 * need a way to "return" the next entry after the end of a list. When
1212 the caller says "before" or "after" we would like to know if he could
1213 cover that interval/threshold or not because this influences the effect
1214 of a newer timestamp of that recentfile. DONE with $opt{info}.
1216 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1218 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1220 loop implies tixu (?). Nope, something like --statefile decides. Per
1221 default we do ...?
1223 T test, I init, X including removals, U nodirtymark
1225 So we have no concept of dirtymarks, we only trust that since we are
1226 running we have observed everything steadily. But people will not let
1227 this program run forever so we must consider both startup penalty and
1228 book keeping for later runs. We keep this for later. For now we write a
1229 long running mirror that merges several intervals.
1231 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1233 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1234 the intervals!
1236 * rersyncrecent, the script: default to one week. The name of the switch
1237 is --after. Other switches? --loop!
1239 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1241 * need a switch --skip-deletes (?)
1243 * need a switch --enduser that tells us that the whole tempfile
1244 discipline is not needed when there is no downstream user. (?)
1246 Without this switch we cannot have a reasonable recent.pl that just
1247 displays the recent additions. Either we accept to download everything.
1248 Or we download temporary files without the typical rsync protocol
1249 advantages.
1251 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1252 File::Temp::tempdir, this might be a win.
1254 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1256 * apropos missing: we have no push, we never know the downstream
1257 servers. People who know their downstream hosts and want to ascertain
1258 something will want additional methods we have never thought about, like
1259 update or delete a certain file.
1261 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1263 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1264 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1265 difference.
1267 * rsync options: it's a bit of a pain that we usually need several rsync
1268 options, like compress, links, times, checksum and that there is no
1269 reasonable default except the original rsync default. I think wee can
1270 safely assume that the rsync options are shared between all recentfile
1271 instances within one recent tree.
1273 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1275 * deletes: if a delete follows an add quickly enough it may happen that
1276 a downstream mirror did not see the add at all! It seems this needs to
1277 be mentioned somewhere. The point here is that even if the downstream is
1278 never missing the principal timeframe it may encounter a "delete" that
1279 has no complimentary "add" anywhere.
1281 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1283 * I suspect the treat of metadata is incorrect during read or something.
1284 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1285 contained more than 6 hours worth of data. At 06:08 we merged into the
1286 1d file. We need to take snapshots of the 6h file over the course of an
1287 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1288 enough.
1290 Much worse: watching the 1h file: right at the moment (at 06:35) it
1291 covers 1218867584-1219120397 which is 70 hours.
1293 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1294 08:19:44 2008, that is when I checked out last time, so it seems to be
1295 aggregating and never truncating?
1297 No, correct is: it is never truncating; but wrong is: it is aggregating.
1298 It does receive a lot of events from time to time from a larger file.
1299 Somehow a large file gets merged into the small one and because the
1300 "meta/merged" attribute is missing, nobody is paying attention. I
1301 believe that I can fix this by making sure that metadata are honoured
1302 during read. DONE and test adjusted.
1304 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1306 * grand renaming plan
1308 remotebase => remoteroot to fit well with localroot DONE
1309 local_path() => localroot seems to me should already work DONE
1310 recentfile_basename => rfilename no need to stress it has no slash DONE
1312 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1313 kern? It would anyway need a deprecation cycle because it is an
1314 important constructor.
1316 * I like the portability that Data::Serializer brings us but the price
1317 is that some day we might find out that it is slowing us a bit. We'll
1318 see.
1320 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1322 * should we not enter the interval of the principal (or the interval of
1323 the merging file?) in every aggregated/merged file?
1325 * we should aim at a first release and give up on thinking about
1326 sanitizing stuff and zloop. Let's just admit that a full traditional
1327 rsync is the only available sanitizer ATM. Otherwise it's complicated
1328 stuff: sanitizing on the origin server, sanitizing on the slaves,
1329 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1330 get the basics out before this becomes a major cause for mess.
1332 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1334 * On OSes not supporting symlinks we expect that RECENT.recent contains
1335 the contents of the principal recentfile. Actually this is identical on
1336 systems supporting symlinks. Simple, what follows from that is that we
1337 need to keep the serializer in the metadata because we cannot read it
1338 from the filename, doesn't it? Of course not. It's a chicken and egg
1339 problem. This leaves us with the problem to actually parse the
1340 serialized data to find out in which format it is. So who can do the 4
1341 or 5 magics we wanted to support? File::LibMagic?
1343 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1345 * remotebase and recentfile_basename are ugly names. Now that we need a
1346 word for the shortest/principal/driving recentfile too we should do
1347 something about it.
1349 localroot is good. rfile is good. local_path() is bad, local_path($path)
1350 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1351 already deprecated.
1353 Up to now remotebase was the string that described the remote root
1354 directory in rsync notation, like pause.perl.org::authors. And
1355 recentfile_basename was "RECENT-1h.yaml".
1357 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1359 * The test that was added in today's checkin is a good start for a test
1360 of rmirror. We should have more methods in Recent.pm: verify,
1361 addmissingfiles. We should verify the current tree, then rmirror it and
1362 then verifytree the copy. We could then add some arbitrary file and let
1363 it be discovered by addmissingfiles, then rmirror again and then
1364 verifytree the copy again.
1366 Then we could start stealing from csync2 sqlite database [no port to
1367 OSX!] and fill a local DB. And methods to compare the database with the
1368 recentfiles. Our strength is that in principle we could maintain state
1369 with a single float. We have synced up to 1234567890.123456. If the Z
1370 file does not add new files all we have to do is mirror the new ones and
1371 delete the goners.
1373 This makes it clear that we should extend current protocol and declare
1374 that we cheat when we add files too late, just to help the other end
1375 keeping track. Ah yes, that's what was meant when zloop was mentioned
1376 earlier.
1378 Maybe need to revisit File::Mirror to help me with this task.
1380 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1382 * There must be an allow-me-to-truncate flag in every recentfile.
1383 Without it one could construct a sequence of updates winning the locking
1384 battle against the aggregator. Only if an aggregator has managed to
1385 merge data over to the next level, truncating can be allowed. DONE with
1386 accessor merged.
1388 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1390 * We should probably guarantee that no duplicates enter the aggregator
1391 array.
1393 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1395 * To get merge operation faster would need a good benchmark test. What
1396 02 spits out isn't reliable enough and is dominated by many other
1397 things. Between
1399 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1400 Author: Andreas J. Koenig <k@k75.(none)>
1401 Date: Sat Aug 2 07:58:04 2008 +0200
1403 and
1405 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1406 Author: Andreas J. Koenig <k@k75.(none)>
1407 Date: Sat Aug 2 11:40:29 2008 +0200
1409 there was a lot of trying but the effect is hardly measurable with
1410 current tests.
1412 * overhead of connecting seems high. When setting
1413 max_files_per_connection to 1 we see that.
1415 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1417 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1419 25138.2215600014/3600 = 6.98283932222261
1421 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1422 be a bug in the aggregator. FIXED
1424 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1426 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1427 we have it, can we mirror it, mirror it!
1429 I fear this needs a new class which might be called
1430 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1431 and treat them as an entity. I realize that a single recentfile may be
1432 sufficient for certain tasks and that it is handy for the low level
1433 programmer but it is not nice to use. If there is a delete in the 1h
1434 file then the 6h file still contains it. Seekers of the best information
1435 need to combine at least some of the recentfiles most of the time.
1437 There is the place for the Z loop!
1439 But the combination is something to collect in a database, isn't it. Did
1440 csync2 just harrumph?
1442 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1444 * it just occurred to me that hosts in the same mirroring pool could
1445 help out each other even without rewriting the recentfile. Just fetch
1446 the stuff to mirror from several places, bingo. But that's something
1447 that should rather live in a separate package or in rsync directly.
1449 * cronjobs are unsuited because with ntp they would all come at the full
1450 minute and disturb each other. Besides that I'd hate to have a backbone
1451 with more than a few seconds latency.
1453 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1455 * a second rsync server with access control for PAUSE. Port? 873 is the
1456 standard port, let's take 8873.
1458 * if there were a filesystem based on this, it would have a slow access
1459 to inexistent files. It would probably provide wrong readdir (only based
1460 on current content) or also a slow one (based on a recentfile written
1461 after the call). But it would provide fast access to existing files. Or
1462 one would deliberately allow slightly blurred answers based on some
1463 sqlite reflection of the recentfiles.
1465 * todo: write a variant of mirror() that combines two or more
1466 recentfiles and treats them like one
1468 * todo: signal handler to remove the tempfile
1470 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1472 * now that we have the symlink I forgot how it should be used in
1473 practice.
1475 * the z loop: add missing files to Z file. Just append them (instead of
1476 prepending). So one guy prepends something from the Y file from time to
1477 time and another guy appends something rather frequently. Collecting
1478 pond. When Y merges into Z, things get epoch and the collecting pond
1479 gets smaller. What exactly are "missing files"?
1481 take note of current epoch of the alpha file, let's call it the
1482 recent-ts
1484 find all files on disk
1486 remove all files registered in the recentworld up to recent-ts
1488 remove all files that have been deleted after recent-ts according to
1489 recentworld
1491 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1493 * rersyncrecent might be a cronjob with a (locked) state file which
1494 contains things like after and maybe last z sync or such?
1496 rrr-mirror might be an alternative name but how would we justify the
1497 three Rs when there is no Re-Rsync-Recent?
1499 With the --loop parameter it is an endless loop, without it is no loop.
1500 At least this is simple.
1502 * todo: new accssor z-interval specifies how often the Z file is updated
1503 against the filesystem. We probably want no epoch stamp on these
1504 entries. And we want to be able to filter the entries (e.g. no
1505 by-modules and by-category tree)
1507 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1509 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1510 into Z. We do not need the other files filled up with filesystem
1511 contents though.
1513 * need interface to query for a file in order to NOT call update on
1514 PAUSE a second time within a short time.
1516 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1518 * recommended update interval? Makes no sense, is different for
1519 different users.
1521 * Moosify
1523 Local Variables:
1524 mode: change-log
1525 change-log-default-name: "Todo"
1526 tab-width: 2
1527 left-margin: 2
1528 End: