| blob | e5caeeab3b02c0662ec00d9431aff2ddc7bf2aee |
1 2010-06-16 Andreas J. Koenig <andk@cpan.org>
3 * rrr-init seems to be sssslllloooowwww. At 3:49 I started it for the
4 video tree. After twelve minutes the file has reached a size of 500k. It
5 is locking and rewriting the YAML after every file. Wasn't there an
6 unlocked variant? For "init" one would not expect locked correctness and
7 a rewrite after every item. Given that this tends to become slower with
8 every file it is really unacceptable.
10 Reminds me of rrr-dirtyupdate (the former name of rrr-update). It seems
11 the code was not dirty enough and that was the reason why it was so
12 slow. And rrr-update does not have a --dirty switch, so it looks like we
13 have no way to force dirty and fast operation yet but we need it for
14 init.
16 Irritating the bootstrapping race: the order {(1) fill initial
17 array, (2) start server} leaves the need for an fsck to cover the time
18 between 1 and 2. Given that regular fsck is needed in any case the price
19 is not high but still doesn't give a warm feeling. Maybe starting the
20 server will by itself do an fsck?
22 2010-06-15 Andreas J. Koenig <andk@cpan.org>
24 * remove the wild alpha alerts and say, it is alpha but parts of it are
25 already working very well or so.
27 2009-12-10 Andreas J. Koenig <andk@cpan.org>
29 * Prominent request for statusfile to safe memory consumption. I'd like
30 to work backwards: get a verification program that verifies how well in
31 sync two nodes are without using frmr itself. Pretend that the forking
32 agent already works and can be interrupted with ^C. Then let it die at
33 random and restart at random and see the memory consumption and the
34 failure rate over time. Compare it with the curve we have now.
36 Of course we also invented the status file as a sharp debugging tool.
37 I'd like to see it in action. See also: F:R:M:Recent::thaw().
39 2009-06-22 Andreas J. Koenig <andk@cpan.org>
41 * Todo: rrr-init to generate some index files so that rrr-server can be
42 started.
44 2009-05-02 Andreas J. Koenig <andk@cpan.org>
46 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
47 so we can use it for normal update with a single file. And add a cronjob
48 that runs rrr-fsck so we get a reminder when something is not working
49 as it should.
51 2009-04-27 Andreas J. Koenig <andk@cpan.org>
53 * Todo: I still see more link_stat errors than expected. I come back to
54 the former suspicion that we do not do necessary bookkeeping on deleted
55 files so that when we reach the "new" event on a meanwhile deleted file
56 we try to mirror it although we could know better. It's just noise, no
57 harm but noise is irritating.
59 * Todo: more experimenting with runstatusfile to get it official
61 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
62 parse the error.
64 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
66 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
68 And that ignore_link_stat_error is propagated in sparse_clone.
70 DONE
72 For the record: the bug was that ignore_link_stat_error was lost during
73 sparse_clone. And then the second bug was that it did not default to
74 true. Both are now fixed.
76 2009-04-25 Andreas J. Koenig <andk@cpan.org>
78 * Todo: fill rrr program with life
80 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
82 * Minor bug: I see complaints about files not existing like
83 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
84 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
86 In the RECENT files these are delete events but of course also new
87 events in older recent files.
89 I think we do not properly remember deletes when we skip-deletes and so
90 do not filter them out when we later see the "new" event.
92 Yes, but that what ignore_link_stat_error is for, we know about this
93 race condition. We must make ignore_link_stat_error default to true.
95 DONE
97 * bug in Done.pm failing to merge two fields into one when a third is
98 present. FIXED and accompanied with new test.
100 2009-04-24 Andreas J. Koenig <andk@cpan.org>
102 * Todo: keep the index clean and run some kind of nondestructive fsck 4
103 times a day.
105 * who is our backbone?
107 http://cpan.cpantesters.org/authors/02STAMP # barbie
108 http://cpan.solfo.com/authors/02STAMP # abh
109 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
111 2009-04-17 Andreas J. Koenig <andk@cpan.org>
113 * the changelog helper I included in the Makefile release memo does not
114 show tags and is much less useful than I thought. gitk is probably much
115 more convenient.
117 2009-04-16 Andreas J. Koenig <andk@cpan.org>
119 * ABH writes:
121 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
122 #### Ydr_.yaml ... DONE
123 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
124 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
125 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
126 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
127 #### gone already at cpan-pause.pl line 0
129 Running without skip-deletes now but cannot reproduce.
131 2009-04-15 Andreas J. Koenig <andk@cpan.org>
133 * consider state directory such we can restart after a ^C or come again
134 from a cronjob.
136 Reuse _runstatusfile: write atomically, see how we can use the dump to
137 start running again. Do not leak and remove globs before writing. Do not
138 only dump one rf, dump the whole complex.
140 * consider the effect of ^C. Is it safe?
142 * want to use Perl::Version to up the versions on every release. Only if
143 things have changed, of course. Does this actually work?
145 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
146 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
148 Must do the bookkeeping myself when this should happen, maybe based on git?
150 I think I prefer the same version number for all pm files. But one also
151 needs a routine for setting them. Ahh:
153 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
155 NEEDSMOREWORK
157 * why indexer wrong?
159 cpan[2]> m /Mirror::Recent/
160 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
161 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
162 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
163 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
164 4 items found
166 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
168 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
170 I just tried MM 6.5102 and it again produces only a provides for Recent,
171 not for the other modules.
173 My own fault. FIXED in the Makefile.PL.
175 2009-04-13 Andreas J. Koenig <andk@cpan.org>
177 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
179 * is it true (as stated at
180 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
181 that rsync can lead to errors when upstream changes before client sync
182 has completed? Can we see the error when we run this on upstream:
184 % perl -e '
185 use Time::HiRes qw(time);
186 while (){
187 open my $fh, ">", "changingfile.txt.new" or die;
188 print $fh (time."\n") x 1000000;
189 close $fh;
190 rename "changingfile.txt.new", "changingfile.txt" or die;
191 }
192 '
194 And on the receiving end:
196 % while true; do
197 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
198 sleep 1;
199 done
201 I cannot get it to fail. Apparently it is sufficient when upstream
202 always writes atomically (which of course is mandatory)?
204 SENT to mailing list Apr 28.
206 * from ABH:
208 > https://fedorahosted.org/InstantMirror/
209 > https://www.redhat.com/mailman/listinfo/instantmirror-list
210 > irc.freenode.net channel #instantmirror
212 A cool name for a project. Inspires me to write a few sentences about
213 bittorrent's role in the grand picture.
215 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
216 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
218 2009-04-12 Andreas J. Koenig <andk@cpan.org>
220 * Interesting last minute bug during real download testing: the output
221 isn't as pretty anymore, it seems that more work is being done than
222 needed.
224 Introducing a highlevel _alluptodate method helped a lot in debugging
225 this.
227 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
228 different on 1h file than on the other files. This lead to frequent
229 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
230 issue.
232 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
234 * Todo: try if we can be faster with a native float package. I'm really
235 glad to have a machine exposing perl floating point bugs, it forced me
236 to retract this item from the changes file: "several speedups in the
237 fakebigfloat code" before I get cpan testers fails.
239 2009-04-11 Andreas J. Koenig <andk@cpan.org>
241 * Bug: when the dirtymark changes upstream, then the downstream server
242 notices it immediately and mirrors the "1h" file but then leaves the
243 rmirror loop and continues with "6h" on the next round through. This
244 behaviour goes on until it reaches the Z file. Because we eat the Z file
245 piecemeal we leave the loop after a while but after that everybody seems
246 to have forgotten that there is some work left to be done. Bad, bad,
247 bad.
249 ETOOCONFUSING better name?
251 Recentfile::get_remote_recentfile_as_tempfile OK
252 Recentfile::resolve_recentfilename split_rfilename DONE
254 Recent::_principal_recentfile_object _principal_recentfile_fromremote
255 Recent::principal_recentfile OK
256 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
258 some void, some not, some not void but called in void context.
260 2009-04-10 Andreas J. Koenig <andk@cpan.org>
262 * Bug? When the dirtymark changes then the second tier hosts quickly
263 reset their DONE state and restart mirroring. But then they mirror
264 potentially outdated index files with inconsistent dirtymark because the
265 first tier box may be a bit behind on the large recentfiles.
267 This means we need a brake when iterating over the recentfiles that
268 refuses to use a recentfile with the wrong dirtymark.
270 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
272 * install the bin/* files and talk about rrr-overview somewhere.
274 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
275 created outside the target tree.
277 * Barbie asks for better logging capabilities
279 2009-03-29 Andreas J. Koenig <andk@cpan.org>
281 * known bug: cannot configure the Recentfile objects to keep delete
282 events. How to call the accessor for that? keep_delete_objects_forever?
284 Sounds acceptable. FIXED (but untested).
286 * known bug: keeps temporary index files lying around.
288 2009-03-24 Andreas J. Koenig <andk@cpan.org>
290 * Some equivalent for
292 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
294 ? DONE
296 2009-03-22 Andreas J. Koenig <andk@cpan.org>
298 * k81 is again client of k75 with different parameters.
300 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
301 about it all the time and nobody reacts accordingly; (2) lots of temp
302 files get created and not removed; culprit the new call to
303 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
304 caller has to remove the tempfile after use.
306 Need the drawing board.
308 PARTLY FIXED: retracting the idea to call
309 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
310 and unseed such that uptodate now means "we have mirrored this
311 recentfile's bunch of files" and seeded means "somebody believes the
312 index file for this rf needs be refreshed". This seems to work now.
314 Still collecting temp files that nobody cares to depollute.
316 2009-03-21 Andreas J. Koenig <andk@cpan.org>
318 * Bug?: should it be harder than it is atm to set the timestamp to the
319 future? YES, FIXED
321 * bug with native integers:
323 -
324 epoch: 997872011
325 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
326 type: new
327 -
328 epoch: 1195248431
329 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
330 type: new
332 Native integer broke when native math was turned off. FIXED
334 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
335 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
336 in this area.
338 Yes, records out of order:
340 447003 -
341 447004 epoch: 995885533
342 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
343 447006 type: new
344 447007 -
345 447008 epoch: 995890358
346 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
347 447010 type: new
348 447011 -
349 447012 epoch: 995892221
350 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
351 447014 type: new
353 FIXED with sanity check and later with the integer fix.
355 * Bug: want the index files in a .recent directory
357 * Bug: lots of dot files are not deleted in time
359 * possible test case: can a delete change the timestamp? This would
360 probably break the order of events.
362 2009-03-20 Andreas J. Koenig <andk@cpan.org>
364 * 1233701831.34486 what's so special about this number/string? It
365 belongs to
366 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
367 atm lives in Y,Q, and Z.
369 It is the first entry after id/--skip-locking which has timestamp
370 1234164228.11325 which represents a file that doesn't exist anymore.
372 * another bug:
374 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
375 _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
376 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
377 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
378 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
379 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
380 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
381 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
382 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
383 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
384 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
385 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
386 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
388 FIXED, it was the "--skip-locking" file where manual intervention was participating
390 * bug on the mirroring slave: when the dirtymark gets increased we
391 probably do not reset the done intervals. The mirrorer stays within
392 tight bounds where it tries to sync with upstream and never seems to
393 finish. In the debugging state file I see lots of identical intervals
394 that do not get collapsed. When I restart the mirrorer it dies with:
396 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
397 _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
398 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
399 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
400 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
401 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
402 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
403 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
404 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
405 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
407 and debugging stands at
409 At: before
410 Brfinterval: Z
411 Ci: 227
412 Dre+1:
413 epoch: 1237400802.5789
414 path: id/J/JO/JOHND/CHECKSUMS
415 type: new
416 Dre-0:
417 epoch: 1237400807.97514
418 path: id/M/MI/MIYAGAWA/CHECKSUMS
419 type: new
420 Dre-1:
421 epoch: 1237400817.94363
422 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
423 type: new
424 Eintervals:
425 -
426 - 900644040
427 - 900644040
428 - []
430 and it is reproducable.
432 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
433 pause has a fresh one.
435 FIXED, it was the third anded term in each of the ifs in the IV block in
436 _register_one: with that we make sure that we do not stamp on valuable
437 interval data.
439 2009-03-17 Andreas J. Koenig <andk@cpan.org>
441 * done: verified the existence of the floating point bug in bleadperl
442 and verified that switching from YAML::Syck to YAML::XS does not resolve
443 it.
445 BTW, the switch was doable with
447 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
449 and should be considered as a separate TODO
451 * todo: integrate a dirty update with two aggregate calls before
452 unlocking for frictionless dirtying DONE
454 * todo: start the second rsync daemon on pause DONE
456 * todo: move index files to .recent: this cannot simply be done by
457 setting filenameroot to .recent/RECENT. Other parts of the modules rely
458 on the fact that dirname(recentfile) is the root of the mirrored tree.
460 2009-03-16 Andreas J. Koenig <andk@cpan.org>
462 * What was the resolution of the mirror.pl delete hook bug? Do we call
463 the delete hook when pause removes a file from MUIR?
465 * Today on pause: Updating 2a13fba..29f284d and installing it for
466 /usr/local/perl-5.10.0{,-RC2}
468 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
469 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
471 2009-03-15 Andreas J. Koenig <andk@cpan.org>
473 * currently recent_events has the side effect of setting dirtymark
474 because it forces a read on the file. That should be transparent, so
475 that the dirtymark call always forces a cache-able(?) read.
477 * The bug below is -- after a lot of trying -- not reproducible on a
478 small script, only in the large test script. The closest to the output
479 below was:
481 #!perl
482 use strict;
483 use Devel::Peek;
484 my $x = "01237123229.8814";
485 my($l,$r);
486 for ($l,$r) {
487 $_ = "x"x34;
488 }
489 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
490 $r = int $r;
491 $l = "1237123231.22458";
492 $r = "1237123231.22458";
493 1 if $l/1.1;
494 Devel::Peek::Dump $l;
495 Devel::Peek::Dump $r;
496 Devel::Peek::Dump $x = $l <=> $r;
497 die "BROKE" if $x;
498 __END__
500 The checked in state at c404a85 fails the test with my
501 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
502 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
503 in the test script always consistent.
505 * Todo: write a test that inserts a second dirty file with an already
506 existing timestamp. DONE
508 * Bug in perl 5.10 on my 64bit box:
510 DB<98> Devel::Peek::Dump $l
511 SV = PVMG(0x19e0450) at 0x142a550
512 REFCNT = 2
513 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
514 IV = 0
515 NV = 1237123231.22458
516 PV = 0x194ce70 "1237123231.22458"\0
517 CUR = 16
518 LEN = 40
520 DB<99> Devel::Peek::Dump $r
521 SV = PVMG(0x19e0240) at 0x142a3e8
522 REFCNT = 2
523 FLAGS = (PADMY,POK,pPOK)
524 IV = 1237123229
525 NV = 1237123229.8814
526 PV = 0x19ff900 "1237123231.22458"\0
527 CUR = 16
528 LEN = 40
530 DB<100> Devel::Peek::Dump $l <=> $r
531 SV = IV(0x19ea6e8) at 0x19ea6f0
532 REFCNT = 1
533 FLAGS = (PADTMP,IOK,pIOK)
534 IV = -1
536 DB<101> Devel::Peek::Dump $l
537 SV = PVMG(0x19e0450) at 0x142a550
538 REFCNT = 2
539 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
540 IV = 0
541 NV = 1237123231.22458
542 PV = 0x194ce70 "1237123231.22458"\0
543 CUR = 16
544 LEN = 40
546 DB<102> Devel::Peek::Dump $r
547 SV = PVMG(0x19e0240) at 0x142a3e8
548 REFCNT = 2
549 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
550 IV = 1237123231
551 NV = 1237123231.22458
552 PV = 0x19ff900 "1237123231.22458"\0
553 CUR = 16
554 LEN = 40
556 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
558 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
560 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
562 * $obj->merge ($other) needs to learn about equal epoch which may happen
563 since dirty_epoch intruded.
565 * Wontfix anytime soon: I think we currently do not support mkdir. Only
566 files!
568 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
570 * Todo: continue working on update(...,$dirty_epoch). It must be
571 followed by a fast_aggregate! DONE
573 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
575 * maybe we need a closest_entry or fitting_interval or something like
576 that. We want to merge an event into the middle of some recentfile.
577 First we do not know which file, then we do not know where to lock,
578 where to enter the new item, when and where to correct the dirtymark.
580 So my thought is we should first find which file.
582 Another part of my brain answers: what would happen if we would enter
583 the new file into the smallest file just like an ordinary new event,
584 just as an old event?
586 (1) we would write a duplicate timestamp? No, this would be easy to
587 avoid
589 (2) we would make the file large quickly? Yes, but so what? We are
590 changing the dirtymark, so are willing to disturb the downstream hosts.
592 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
594 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
596 leak!
598 https://rt.cpan.org/Ticket/Display.html?id=41199
600 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
602 Strangest output being HAKANARDO who managed to upload
604 Here is a better oneliner that includes also the first line of each
605 finding:
607 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
609 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
611 * memory leak in the syncher? It currently weighs 100M.
613 Update 2008-11-02:
615 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
618 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
620 * lookup by epoch and by path and use this ability on the pause to never
621 again register a file twice that doesn't need it. Let's call it
622 contains().
624 * after the dirtymark is done: fill up recentfiles with fake (historic)
625 entries; fill up with individual corrections; algorithm maybe to be done
626 with bigfloat so that we can always place something in the middle
627 between two entries. Before we must switch to bigfloat we could try to
628 use Data::Float::nextup to get the.
630 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
631 directly.
633 * dirtymark now lives in Recentfile, needs to be used in rmirror.
635 * find out why the downloader died after a couple of hours without a net
636 connection. Write a test that survives the not-existence of the other
637 end forever.
639 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
641 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
642 it makes less sense with tight coupling for secondary files. But asking
643 the server what the current epoch is might be cheaper on HTTP than on
644 rsync. (Needs to be evaluated)
646 * remove the 0.00 from the verbose overview in the Merged column in the
647 Z row. DONE
649 * write tests that expose the problems of the last few days: cascading
650 client/server roles, tight coupling for secondary RFs, deletes after
651 copies.
653 * Some day we might want to have policy options for the slave:
654 tight/loose/no coupling with upstream for secondary RFs. tight is what
655 we have now. loose would wait until a gap occurs that can be closed.
657 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
659 * revisit all $rfs->[$i+1] places if they now make sense still
661 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
663 * another bug is the fact that the mirror command deletes files before
664 it unhides the index file, thus confusing downstream slaves. We must not
665 delete before unhiding and must delete after unhiding. FIXED.
667 * new complication about the slave that is playing a server role.
668 Currently we mirror from newest to oldest with a hidden temporary file
669 as index. And when one file is finished, we unhide the index file.
670 Imagine the cascading server/slave is dead for a day. It then starts
671 mirroring again with the freshest thing and unhides the freshest index
672 file when it has worked through it. In that moment it exposes a time
673 hole. Because it now works on the second recentfile which is still
674 hidden.
676 We currently do nothing special to converge after such a drop out. At
677 least not intentionally and robustly and thought through.
679 The algorithm we use to seed the next file needs quite a lot of more
680 robustness than it currently has. Something to do with looking at the
681 merged element of the next rf and when it has dropped off, we seed
682 immediately. And if it ramains dropped off, we seed again, of course.
684 Nope, looking from smaller to larger RFS we look at the merged element
685 of this RF and at the minmax/max element of the next RF. If that
686 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
687 seeding it.
689 And we need a public accessor seed and unseed or seeded. But not the mix
690 of public and private stuff that then is used behind the back.
692 And then the secondary* stuff must go.
694 And we must understand what the impact is on the DONE system. Can it go
695 unnoticed that there was a hole? And could the DONE system have decided
696 the hole is covered? This should be testable with three directories where
697 the middle stops working for a while. Done->merge is suspicious, we must
698 stop it from merging non-conflatable neighbors due to broken continuity.
700 FIXED
702 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
704 * Slaven suggests to have the current epoch or the whole current
705 recentfile available from the HTTP server and take it away with
706 keepalive. This direction goes the granularity down to subseconds.
708 We might want to rewrite everything to factor out transport and allow
709 the whole thing to run via HTTP.
711 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
713 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
714 symlink?
716 * smoker on k81 fetching from k75 to verify cascading works. See
717 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
719 * maybe the loop should wait for CHECKSUMS file after every upload. And
720 CPAN.pm needs to deal with timestamps in the future.
722 * do not forget the dirtymark!
724 Text: have a new flag on recentfiles with the meaning: if this
725 changes, you're required to run a full rsync over all the files. The
726 reason why we set it would probably be: some foul happened. we injected
727 files in arbitrary places or didn't inject them although they changed.
728 The content of the flag? Timestamp? The relation between the
729 recentfiles would have to be inheritance from the principal, because any
730 out of band changes would soon later propagate to the next recentfile.
732 By upping the flag often one can easily ruin the slaves.
734 last out of band change? dirtymark?
736 Anyway, this implies that we read a potentially existing recentfile
737 before we write one.
739 And it implies that we have an eventloop that keeps us busy in 2-3
740 cycles, one for current stuff (tight loop) and one for the recentfiles
741 (cascade when principal has changed), one for the old stuff after a
742 dirtymark change.
744 And it implies that the out-of-band change in any of the recentfiles
745 must have a lock on the principal file and there is the place to set the
746 dirtymark.
748 * start a FAQ, especially quick start guide questions. Also to aid those
749 problematic areas where we have no good solution, like the "links"
750 option to rsync.
752 * wish feedback when we are slow.
754 * reduce mccabe
756 * Remove a few DEBUG statements.
758 * The multiple-rrr way of doing things needs a new option to rmirror,
759 like piecemeal or so. Not urgent because after the first pass through,
760 things run smoothely. It's only ugly during the first pass.
762 * I have the suspicion that the code is broken that decides if the
763 neighboring RF needs to be seeded. I fear when too much time has gone
764 between two calls (in our case more than one hour), it would not seed
765 the neighbor. Of course this will never be noticed, so we need a good
766 test for it.
768 * local/localroot confusion: I currently pass both options but one must
769 do.
771 * accounts for early birds on PAUSE rsync daemon.
773 * hardcoded 20 seconds
775 * who mirrors the index? DOING now.
777 * which CPAN mirrors offer rsync?
779 * visit all XXX, visit all _float places
781 * rename the pathdb stuff, it's too confusing. No idea how.
783 * rrr-inotify, backpan, rrr-register
785 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
787 * current bugs: the pathdb seems to get no reset, the seeding of the
788 secondaryttl stuff seems not to have an effect. Have helped myself with
789 a rand(10), need to fix this back. So not checked in. Does the rand
790 thing even help?
792 The rand thing helps. The secondaryttl stuff was in the wrong line,
793 fixed now.
795 The pathdb stuff was because I called either _pathdb or __pathdb on the
796 wrong object. FIXED now.
798 * It's not so beautiful if we never fetch the recentfiles that are not
799 the principal, even if this is correct behaviour. We really do not need
800 them after we have fetched the whole content.
802 OK, we want a switch for that: secondaryttl DONE
804 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
806 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
807 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
808 it lets 40 follow. FIXED
810 * See that the long running process really only updates the principal
811 file unless it has missed a timespan during which something happened. If
812 nothing happened, it must notice even when it misses the timespan. DONE
814 * we must throw away the pathdb when we have reached the end of Z. From
815 that moment we can have a very small pathdb because the only reason for
816 a pathdb is that we know to ignore old records in old files. We won't
817 need this pathdb again before the next full pass over the data is
818 necessary and then we will rebuild it as we go along. DONE
820 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
822 * I think, Done::register_one is doing wrong in that it does not
823 conflate neighboring pieces. The covered() method cannot do this because
824 it has no recent_events array at hand. But register_one has it and could
825 do it and for some reason misses to do it (sometimes).
827 This means that the three tests I just wrote can probably not survive
828 because they test with an already broken Done structure.
830 The art now is to detect how it happens, then to reproduce, then write a
831 test, then fix it.
833 So from the logfile this is what happens: we have a good interval with
834 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
835 top of it. Locally we now mirror F2 and open a new done interval for it.
836 Then we mirror F1 but this time with the timestamp T1b. And when we then
837 try to close the gap, we do not find T1 but instead something older. We
838 should gladly accept this older piece and this would fix this bug.
840 FIXED
842 * bug to fix: when the 1h file changes while rmirror is running, we do
843 correctly sync the new files but never switch to the 6h file but rather
844 stay in a rather quick loop that fetches the 1h file again and again.
846 Is it possible that we initialize a new object? Or does
847 get_remote_recentfile_as_tempfile overwrite something in myself?
849 Want a new option: _runstatusfile => $file which frequently dumps the
850 state of all recentfiles to a file.
852 FIXED
854 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
856 * Todo: now teach update to verify the timestamp is about to write
857 against the previous and use _increase_a_bit if it doesn't comply with
858 strict monotony. DONE
860 * The problem of rounding. So far perl's default precision was
861 sufficient. One day it won't be. FakeFloat has an easy job when it is
862 only reading and other machines have written correctly. But when we want
863 to write a floating point number that is a bit larger than the other
864 one, then we need our own idea of precision.
866 Slaven said: just append a "1". This might be going towards the end of
867 usability too quickly. I'd like something that actually uses the decimal
868 system. Well, appending a 1 also does this but...
870 E.g. we have 1.0. nextup on this architecture is starting with
871 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
872 taken the 1.0000000000000003 and the next user comes and the time tells
873 him 1.0 again. He has to beat my number without stepping over the
874 nextup. This is much less space than I had when I chose 1,2,3.
876 What is also irritating is that nextup is architecture dependent. The
877 128 bit guy must choose very long numbers to fit in between whereas the
878 other one with 16 bit uses larger steps. But then the algorithm is the
879 same for both, so that would be a nice thing.
881 I see two situation where we need this. One is when Time::HiRes returns
882 us a value that is <= the last entry in our recentfile. In this case
883 (let's call it the end-case) we must fill the region between that number
884 and the next higher native floating point number. The other is when we
885 inject an old file into an old recentfile (we would then also set a new
886 dirtymark). We find the integer value already taken and need a slightly
887 different one (let's call it the middle-case). The difference between
888 the two situations is that the next user will want to find something
889 higher than my number in the end-case and something lower than my number
890 in the middle case.
892 So I suggest we give the function both a value and an upper bound and it
893 calculates us a primitive middle. The upper bound in the middle-case is
894 the next integer. The upper bound on the end-case is the nextup floating
895 point number. But the latter poses another problem: if we have occupied
896 the middle m between x and nextup(x), then the nextup(m) will probably
897 not be the same as nextup(x) because some rounding will take place
898 before the nextup is calculated and when the rounding reaches the
899 nextup(x), we will end up at nextup(nextup(x)).
901 So we really need to consider the nextup and the nextdown from there and
902 then the middle and that's the number we may approach asymptotically.
903 Ugly. But DONE.
905 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
907 * consider deprecating the use of RECENT.recent as a symlink. It turns
908 out to need extra hoops with the rsync options and just isn't worth it.
909 Or maybe these extra hoops are needed anyway for the rest of the tree?
910 Nope, can't be the case because not all filesystems support symlinks.
912 But before doing the large step, I'll deprecate the call of
913 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
914 only introduced to resolve RECENT.recent and complicates the routine far
915 beyond what it deserves.
917 DONE. Won't deprecate RECENT.recent, just moved its handling to the
918 supervisor.
920 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
922 * I think it's a bug that the rsync_option links must be set to true in
923 order to support RECENT.recent and that nobody cares to set it
924 automatically. Similar for ignore_link_stat_errors.
926 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
928 * Todo: find all todos together and make a plan what is missing for a
929 release.
931 - verifytree or something like that. fsck maybe.
933 - rersyncrecent, the script itself? What it do?
935 - a way to only mirror the recentfiles without mirroring the whole
936 remote system such that people can decide to mirror only partially see
937 also 2008-08-30. .shadow-xxx directory? this also needed for a
938 filesystem that is still incomplete and might need the mirrorfiles for
939 lookup(?)
941 - long living objects that mirror again and again. Inject something
942 into ta, see how it goes over to tb.
944 - how do we continue filling up the DONE system when we use an object
945 for the second time? "fully covered" and "uptodate" or new terminology.
947 - overview called on the wrong file should be understandable
949 - the meta data field that must change when we fake something up so that
950 the downstream people know they have to re-fetch everything.
952 - how tolerant are we against missing files upstream? how do we keep
953 track? there are legitimate cases where we did read upstream index right
954 before a file got deleted there and then find that file as new and want
955 it. There are other cases that are not self healing and must be tracked
956 and bugreported.
958 - how, exactly, do we have to deal with deletes? With rsync errors?
960 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
961 authors) failed: No such file or directory (2)
963 The file above is a delete in 1h and a new in file 1M and the
964 delete in the locally running rmirror did not get propagated to the 1M
965 object. Bug. And the consequence is a standstill.
967 It seems that a slave that works with a file below the principal needs
968 to merge things all the way up to get rid of later deletes. Or keep
969 track of all deletes and skip them later. So we need a trackdeletes.pm
970 similar to the done.pm?
972 see also 2008-08-20 about spurious deletes that really have no add
973 counterpart and yet they are not wrong.
975 - consider the effect when resyncing the recentfile takes longer than
976 the time per loop. Then we never rsync any file. We need to diagnose
977 that and force an increase of that loop time. But when we later are fast
978 enough again because the net has recovered, then we need to switch back
979 to original parameters. ERm, no, it's enough to keep syncing at least
980 one file before refetching an index file.
982 - remember to verify that no temp files are left lying around and the
983 signal handler
985 - status file for not long running jobs that want to track upstream with
986 a, say, cronjob.
988 - revisit all XXX _float areas and study Sub::Exporter DONE
990 - persistent DB even though we just said we do not need it. Just for
991 extended capabilities and time savings when, for example, upstream
992 announces a reset and we get new recentfiles and could then limit
993 ourselves to a subset of files (those that have a changed epoch) in a
994 first pass and would only then do the loop to verify the rest. Or
995 something.
997 * Todo: aggregate files should know their feed and finding the principal
998 should be done stepwise. (?)
1000 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
1001 left locks around. DONE
1003 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1005 * maybe extend the _overview so that it always says if and where the
1006 last file is in the next file and where the next event in the next rf
1007 would lie. No, don't like this anymore. REJECT
1009 * take the two new redundant tests out again, only the third must
1010 survive. DONE
1012 * Todo: add a sanity check if the merged structure is really pointing to
1013 a different rf and that this different rf is larger. DONE
1015 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1017 * now test, if they are overlapping. And test if there is a file in the
1018 next rf that would fit into this rf's interval.
1020 1h 1222324012.8474 1222322541.7963 0.4086
1021 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
1022 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
1023 1W 1222313218.3626 1221708477.5829 167.9835
1025 I suspect that somebody writes a merged timestamp without having merged
1026 and then somebody else relies on it.
1028 If aggregate is running, the intervals must not be extravagated, if it
1029 is not running, there must not be bounds, the total number of events in
1030 the system must be counted and must be controlled throughout the tests.
1031 That the test required the additional update was probably nonsense,
1032 because aggregate can cut pieces too. FIXED & DONE
1034 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1036 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1037 changed. FIXED
1039 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1041 * Most apparent bug at the moment is that the recentfiles are fetched
1042 too often. Only the principal should be fetched and if it has not
1043 changed, the others should not be refetched. ATM I must admit that I'm
1044 happy that we refetch more often than needed because I can more easily
1045 fix bugs while the thing is running.
1047 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1048 done but the done system does not know about it. The reason for the
1049 failure is not known and we never reach the status uptodate because of
1050 this. We must get over it.
1052 Later it turns out that the origin server had a bug somewhere.
1053 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1054 array of the recentfile one position above. The bug was my own.
1056 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1058 * There is the race condition where the server does a delete and the
1059 slave does not yet know and then tries to download it because he sees
1060 the new. So for this time window we must be more tolerant against
1061 failure. If we cannot download a file, we should just skip it and should
1062 not retry immediately. The whole system should discover the lost thing
1063 later. Keeping track with the DONE system should really be a no brainer.
1065 But there is something more: the whole filesystem is a database and the
1066 recentfiles are one possible representation of it. It's a pretty useful
1067 representation I think that's why I have implemented something around
1068 it. But for strictly local operation it has little value. For local
1069 operation we would much rather have a database. So we would enter every
1070 recentfile reading and every rsync operation and for every file the last
1071 state change and what it leads to. Then we would always ignore older
1072 records without the efforts involved with recentfiles.
1074 The database would have: path,recentepoch,rsyncedon,deletedon
1076 Oh well, not yet clear where this leads to.
1078 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1080 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1081 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1083 * statusfile, maybe only on demand, alone to have a sharp debugging
1084 tool. It is locked and all recentfiles dump themselves into it and we
1085 can build a viewer that lets us know where we stand and what's inside.
1087 * remember: only the principal recentfile needs expiration, all others
1088 shall be expired by principal if it discovers that something has move
1089 upstream.
1091 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1093 * Always check if we stringify to a higher value than in the entry
1094 before. DONE
1096 * And in covered make an additional check if we would be able to see a
1097 numerical difference between the two numbers and if we can't then switch
1098 to a different, more expensive algorithm. Do not want to be caught by
1099 floating surprises. DONE
1101 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1103 * caching has several aspects here: we can cache the interval of the
1104 recentfile which only will change when the mtime of the file changes. We
1105 must re-mirror the recentfile when its ttl has expired. Does have_read
1106 tell you anything? It counts nothing at all. Only the mtime is
1107 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1108 thing the have_mirrored because it is unrelated to the mtime.
1110 * Robustness of floating point calculations! I always thought that the
1111 string calculated by the origin server for the floating representation
1112 of the epoch time is just a string. When we convert it to a number and
1113 later back to a string, the other computer might come to a different
1114 conclusion. This must not happen, we want to preserve it under any
1115 circumstances. I will have to write tests with overlong sequences that
1116 get lost in arithmetic and must see if all still works well. DONE
1118 But one fragile point remains: if one host considers a>b and the other
1119 one considers them == but no eq. To prevent this, we must probably do
1120 some extra homework. DONE
1122 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1124 * the concept of tracking DONE needs an object per recentfile that has
1125 something like these methods:
1127 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1129 covered() register() covered()
1131 The unclear thing is how we translate points in time into intervals. We
1132 could pass a reference to the current recent_events array when running
1133 we_have(xxx) and let the DONE object iterate over it such that it only
1134 has to store a list of intervals that can melt into each other. Ah, even
1135 passing the list together with a list of indexes seems feasiable.
1137 Or maybe ask for the inverted list?
1139 Whenever the complete array is covered by the interval we say we are
1140 fully covered and if the recentfile is not expired, we are uptodate.
1142 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1144 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1146 * need a way to "return" the next entry after the end of a list. When
1147 the caller says "before" or "after" we would like to know if he could
1148 cover that interval/threshold or not because this influences the effect
1149 of a newer timestamp of that recentfile. DONE with $opt{info}.
1151 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1153 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1155 loop implies tixu (?). Nope, something like --statefile decides. Per
1156 default we do ...?
1158 T test, I init, X including removals, U nodirtymark
1160 So we have no concept of dirtymarks, we only trust that since we are
1161 running we have observed everything steadily. But people will not let
1162 this program run forever so we must consider both startup penalty and
1163 book keeping for later runs. We keep this for later. For now we write a
1164 long running mirror that merges several intervals.
1166 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1168 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1169 the intervals!
1171 * rersyncrecent, the script: default to one week. The name of the switch
1172 is --after. Other switches? --loop!
1174 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1176 * need a switch --skip-deletes (?)
1178 * need a switch --enduser that tells us that the whole tempfile
1179 discipline is not needed when there is no downstream user. (?)
1181 Without this switch we cannot have a reasonable recent.pl that just
1182 displays the recent additions. Either we accept to download everything.
1183 Or we download temporary files without the typical rsync protocol
1184 advantages.
1186 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1187 File::Temp::tempdir, this might be a win.
1189 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1191 * apropos missing: we have no push, we never know the downstream
1192 servers. People who know their downstream hosts and want to ascertain
1193 something will want additional methods we have never thought about, like
1194 update or delete a certain file.
1196 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1198 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1199 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1200 difference.
1202 * rsync options: it's a bit of a pain that we usually need several rsync
1203 options, like compress, links, times, checksum and that there is no
1204 reasonable default except the original rsync default. I think wee can
1205 safely assume that the rsync options are shared between all recentfile
1206 instances within one recent tree.
1208 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1210 * deletes: if a delete follows an add quickly enough it may happen that
1211 a downstream mirror did not see the add at all! It seems this needs to
1212 be mentioned somewhere. The point here is that even if the downstream is
1213 never missing the principal timeframe it may encounter a "delete" that
1214 has no complimentary "add" anywhere.
1216 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1218 * I suspect the treat of metadata is incorrect during read or something.
1219 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1220 contained more than 6 hours worth of data. At 06:08 we merged into the
1221 1d file. We need to take snapshots of the 6h file over the course of an
1222 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1223 enough.
1225 Much worse: watching the 1h file: right at the moment (at 06:35) it
1226 covers 1218867584-1219120397 which is 70 hours.
1228 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1229 08:19:44 2008, that is when I checked out last time, so it seems to be
1230 aggregating and never truncating?
1232 No, correct is: it is never truncating; but wrong is: it is aggregating.
1233 It does receive a lot of events from time to time from a larger file.
1234 Somehow a large file gets merged into the small one and because the
1235 "meta/merged" attribute is missing, nobody is paying attention. I
1236 believe that I can fix this by making sure that metadata are honoured
1237 during read. DONE and test adjusted.
1239 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1241 * grand renaming plan
1243 remotebase => remoteroot to fit well with localroot DONE
1244 local_path() => localroot seems to me should already work DONE
1245 recentfile_basename => rfilename no need to stress it has no slash DONE
1247 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1248 kern? It would anyway need a deprecation cycle because it is an
1249 important constructor.
1251 * I like the portability that Data::Serializer brings us but the price
1252 is that some day we might find out that it is slowing us a bit. We'll
1253 see.
1255 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1257 * should we not enter the interval of the principal (or the interval of
1258 the merging file?) in every aggregated/merged file?
1260 * we should aim at a first release and give up on thinking about
1261 sanitizing stuff and zloop. Let's just admit that a full traditional
1262 rsync is the only available sanitizer ATM. Otherwise it's complicated
1263 stuff: sanitizing on the origin server, sanitizing on the slaves,
1264 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1265 get the basics out before this becomes a major cause for mess.
1267 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1269 * On OSes not supporting symlinks we expect that RECENT.recent contains
1270 the contents of the principal recentfile. Actually this is identical on
1271 systems supporting symlinks. Simple, what follows from that is that we
1272 need to keep the serializer in the metadata because we cannot read it
1273 from the filename, doesn't it? Of course not. It's a chicken and egg
1274 problem. This leaves us with the problem to actually parse the
1275 serialized data to find out in which format it is. So who can do the 4
1276 or 5 magics we wanted to support? File::LibMagic?
1278 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1280 * remotebase and recentfile_basename are ugly names. Now that we need a
1281 word for the shortest/principal/driving recentfile too we should do
1282 something about it.
1284 localroot is good. rfile is good. local_path() is bad, local_path($path)
1285 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1286 already deprecated.
1288 Up to now remotebase was the string that described the remote root
1289 directory in rsync notation, like pause.perl.org::authors. And
1290 recentfile_basename was "RECENT-1h.yaml".
1292 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1294 * The test that was added in today's checkin is a good start for a test
1295 of rmirror. We should have more methods in Recent.pm: verify,
1296 addmissingfiles. We should verify the current tree, then rmirror it and
1297 then verifytree the copy. We could then add some arbitrary file and let
1298 it be discovered by addmissingfiles, then rmirror again and then
1299 verifytree the copy again.
1301 Then we could start stealing from csync2 sqlite database [no port to
1302 OSX!] and fill a local DB. And methods to compare the database with the
1303 recentfiles. Our strength is that in principle we could maintain state
1304 with a single float. We have synced up to 1234567890.123456. If the Z
1305 file does not add new files all we have to do is mirror the new ones and
1306 delete the goners.
1308 This makes it clear that we should extend current protocol and declare
1309 that we cheat when we add files too late, just to help the other end
1310 keeping track. Ah yes, that's what was meant when zloop was mentioned
1311 earlier.
1313 Maybe need to revisit File::Mirror to help me with this task.
1315 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1317 * There must be an allow-me-to-truncate flag in every recentfile.
1318 Without it one could construct a sequence of updates winning the locking
1319 battle against the aggregator. Only if an aggregator has managed to
1320 merge data over to the next level, truncating can be allowed. DONE with
1321 accessor merged.
1323 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1325 * We should probably guarantee that no duplicates enter the aggregator
1326 array.
1328 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1330 * To get merge operation faster would need a good benchmark test. What
1331 02 spits out isn't reliable enough and is dominated by many other
1332 things. Between
1334 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1335 Author: Andreas J. Koenig <k@k75.(none)>
1336 Date: Sat Aug 2 07:58:04 2008 +0200
1338 and
1340 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1341 Author: Andreas J. Koenig <k@k75.(none)>
1342 Date: Sat Aug 2 11:40:29 2008 +0200
1344 there was a lot of trying but the effect is hardly measurable with
1345 current tests.
1347 * overhead of connecting seems high. When setting
1348 max_files_per_connection to 1 we see that.
1350 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1352 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1354 25138.2215600014/3600 = 6.98283932222261
1356 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1357 be a bug in the aggregator. FIXED
1359 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1361 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1362 we have it, can we mirror it, mirror it!
1364 I fear this needs a new class which might be called
1365 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1366 and treat them as an entity. I realize that a single recentfile may be
1367 sufficient for certain tasks and that it is handy for the low level
1368 programmer but it is not nice to use. If there is a delete in the 1h
1369 file then the 6h file still contains it. Seekers of the best information
1370 need to combine at least some of the recentfiles most of the time.
1372 There is the place for the Z loop!
1374 But the combination is something to collect in a database, isn't it. Did
1375 csync2 just harrumph?
1377 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1379 * it just occurred to me that hosts in the same mirroring pool could
1380 help out each other even without rewriting the recentfile. Just fetch
1381 the stuff to mirror from several places, bingo. But that's something
1382 that should rather live in a separate package or in rsync directly.
1384 * cronjobs are unsuited because with ntp they would all come at the full
1385 minute and disturb each other. Besides that I'd hate to have a backbone
1386 with more than a few seconds latency.
1388 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1390 * a second rsync server with access control for PAUSE. Port? 873 is the
1391 standard port, let's take 8873.
1393 * if there were a filesystem based on this, it would have a slow access
1394 to inexistent files. It would probably provide wrong readdir (only based
1395 on current content) or also a slow one (based on a recentfile written
1396 after the call). But it would provide fast access to existing files. Or
1397 one would deliberately allow slightly blurred answers based on some
1398 sqlite reflection of the recentfiles.
1400 * todo: write a variant of mirror() that combines two or more
1401 recentfiles and treats them like one
1403 * todo: signal handler to remove the tempfile
1405 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1407 * now that we have the symlink I forgot how it should be used in
1408 practice.
1410 * the z loop: add missing files to Z file. Just append them (instead of
1411 prepending). So one guy prepends something from the Y file from time to
1412 time and another guy appends something rather frequently. Collecting
1413 pond. When Y merges into Z, things get epoch and the collecting pond
1414 gets smaller. What exactly are "missing files"?
1416 take note of current epoch of the alpha file, let's call it the
1417 recent-ts
1419 find all files on disk
1421 remove all files registered in the recentworld up to recent-ts
1423 remove all files that have been deleted after recent-ts according to
1424 recentworld
1426 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1428 * rersyncrecent might be a cronjob with a (locked) state file which
1429 contains things like after and maybe last z sync or such?
1431 rrr-mirror might be an alternative name but how would we justify the
1432 three Rs when there is no Re-Rsync-Recent?
1434 With the --loop parameter it is an endless loop, without it is no loop.
1435 At least this is simple.
1437 * todo: new accssor z-interval specifies how often the Z file is updated
1438 against the filesystem. We probably want no epoch stamp on these
1439 entries. And we want to be able to filter the entries (e.g. no
1440 by-modules and by-category tree)
1442 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1444 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1445 into Z. We do not need the other files filled up with filesystem
1446 contents though.
1448 * need interface to query for a file in order to NOT call update on
1449 PAUSE a second time within a short time.
1451 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1453 * recommended update interval? Makes no sense, is different for
1454 different users.
1456 * Moosify
1458 Local Variables:
1459 mode: change-log
1460 change-log-default-name: "Todo"
1461 tab-width: 2
1462 left-margin: 2
1463 End: