| blob | 3fb437e73d5d67ddca51321b6372d4aec3f1db33 |
1 2010-07-06 Andreas J. Koenig <andk@cpan.org>
3 * rrr-init now being fast we can now remove the alpha alerts. Done.
5 2010-06-16 Andreas J. Koenig <andk@cpan.org>
7 * rrr-init seems to be sssslllloooowwww. At 3:49 I started it for the
8 video tree. After twelve minutes the file has reached a size of 500k. It
9 is locking and rewriting the YAML after every file. Wasn't there an
10 unlocked variant? For "init" one would not expect locked correctness and
11 a rewrite after every item. Given that this tends to become slower with
12 every file it is really unacceptable.
14 Reminds me of rrr-dirtyupdate (the former name of rrr-update). It seems
15 the code was not dirty enough and that was the reason why it was so
16 slow. And rrr-update does not have a --dirty switch, so it looks like we
17 have no way to force dirty and fast operation yet but we need it for
18 init.
20 Irritating the bootstrapping race: the order {(1) fill initial
21 array, (2) start server} leaves the need for an fsck to cover the time
22 between 1 and 2. Given that regular fsck is needed in any case the price
23 is not high but still doesn't give a warm feeling. Maybe starting the
24 server will by itself do an fsck?
26 2010-06-15 Andreas J. Koenig <andk@cpan.org>
28 * remove the wild alpha alerts and say, it is alpha but parts of it are
29 already working very well or so.
31 2009-12-10 Andreas J. Koenig <andk@cpan.org>
33 * Prominent request for statusfile to safe memory consumption. I'd like
34 to work backwards: get a verification program that verifies how well in
35 sync two nodes are without using frmr itself. Pretend that the forking
36 agent already works and can be interrupted with ^C. Then let it die at
37 random and restart at random and see the memory consumption and the
38 failure rate over time. Compare it with the curve we have now.
40 Of course we also invented the status file as a sharp debugging tool.
41 I'd like to see it in action. See also: F:R:M:Recent::thaw().
43 2009-06-22 Andreas J. Koenig <andk@cpan.org>
45 * Todo: rrr-init to generate some index files so that rrr-server can be
46 started.
48 2009-05-02 Andreas J. Koenig <andk@cpan.org>
50 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
51 so we can use it for normal update with a single file. And add a cronjob
52 that runs rrr-fsck so we get a reminder when something is not working
53 as it should.
55 2009-04-27 Andreas J. Koenig <andk@cpan.org>
57 * Todo: I still see more link_stat errors than expected. I come back to
58 the former suspicion that we do not do necessary bookkeeping on deleted
59 files so that when we reach the "new" event on a meanwhile deleted file
60 we try to mirror it although we could know better. It's just noise, no
61 harm but noise is irritating.
63 * Todo: more experimenting with runstatusfile to get it official
65 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
66 parse the error.
68 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
70 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
72 And that ignore_link_stat_error is propagated in sparse_clone.
74 DONE
76 For the record: the bug was that ignore_link_stat_error was lost during
77 sparse_clone. And then the second bug was that it did not default to
78 true. Both are now fixed.
80 2009-04-25 Andreas J. Koenig <andk@cpan.org>
82 * Todo: fill rrr program with life
84 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
86 * Minor bug: I see complaints about files not existing like
87 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
88 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
90 In the RECENT files these are delete events but of course also new
91 events in older recent files.
93 I think we do not properly remember deletes when we skip-deletes and so
94 do not filter them out when we later see the "new" event.
96 Yes, but that what ignore_link_stat_error is for, we know about this
97 race condition. We must make ignore_link_stat_error default to true.
99 DONE
101 * bug in Done.pm failing to merge two fields into one when a third is
102 present. FIXED and accompanied with new test.
104 2009-04-24 Andreas J. Koenig <andk@cpan.org>
106 * Todo: keep the index clean and run some kind of nondestructive fsck 4
107 times a day.
109 * who is our backbone?
111 http://cpan.cpantesters.org/authors/02STAMP # barbie
112 http://cpan.solfo.com/authors/02STAMP # abh
113 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
115 2009-04-17 Andreas J. Koenig <andk@cpan.org>
117 * the changelog helper I included in the Makefile release memo does not
118 show tags and is much less useful than I thought. gitk is probably much
119 more convenient.
121 2009-04-16 Andreas J. Koenig <andk@cpan.org>
123 * ABH writes:
125 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
126 #### Ydr_.yaml ... DONE
127 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
128 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
129 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
130 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
131 #### gone already at cpan-pause.pl line 0
133 Running without skip-deletes now but cannot reproduce.
135 2009-04-15 Andreas J. Koenig <andk@cpan.org>
137 * consider state directory such we can restart after a ^C or come again
138 from a cronjob.
140 Reuse _runstatusfile: write atomically, see how we can use the dump to
141 start running again. Do not leak and remove globs before writing. Do not
142 only dump one rf, dump the whole complex.
144 * consider the effect of ^C. Is it safe?
146 * want to use Perl::Version to up the versions on every release. Only if
147 things have changed, of course. Does this actually work?
149 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
150 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
152 Must do the bookkeeping myself when this should happen, maybe based on git?
154 I think I prefer the same version number for all pm files. But one also
155 needs a routine for setting them. Ahh:
157 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
159 NEEDSMOREWORK
161 * why indexer wrong?
163 cpan[2]> m /Mirror::Recent/
164 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
165 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
166 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
167 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
168 4 items found
170 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
172 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
174 I just tried MM 6.5102 and it again produces only a provides for Recent,
175 not for the other modules.
177 My own fault. FIXED in the Makefile.PL.
179 2009-04-13 Andreas J. Koenig <andk@cpan.org>
181 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
183 * is it true (as stated at
184 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
185 that rsync can lead to errors when upstream changes before client sync
186 has completed? Can we see the error when we run this on upstream:
188 % perl -e '
189 use Time::HiRes qw(time);
190 while (){
191 open my $fh, ">", "changingfile.txt.new" or die;
192 print $fh (time."\n") x 1000000;
193 close $fh;
194 rename "changingfile.txt.new", "changingfile.txt" or die;
195 }
196 '
198 And on the receiving end:
200 % while true; do
201 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
202 sleep 1;
203 done
205 I cannot get it to fail. Apparently it is sufficient when upstream
206 always writes atomically (which of course is mandatory)?
208 SENT to mailing list Apr 28.
210 * from ABH:
212 > https://fedorahosted.org/InstantMirror/
213 > https://www.redhat.com/mailman/listinfo/instantmirror-list
214 > irc.freenode.net channel #instantmirror
216 A cool name for a project. Inspires me to write a few sentences about
217 bittorrent's role in the grand picture.
219 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
220 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
222 2009-04-12 Andreas J. Koenig <andk@cpan.org>
224 * Interesting last minute bug during real download testing: the output
225 isn't as pretty anymore, it seems that more work is being done than
226 needed.
228 Introducing a highlevel _alluptodate method helped a lot in debugging
229 this.
231 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
232 different on 1h file than on the other files. This lead to frequent
233 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
234 issue.
236 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
238 * Todo: try if we can be faster with a native float package. I'm really
239 glad to have a machine exposing perl floating point bugs, it forced me
240 to retract this item from the changes file: "several speedups in the
241 fakebigfloat code" before I get cpan testers fails.
243 2009-04-11 Andreas J. Koenig <andk@cpan.org>
245 * Bug: when the dirtymark changes upstream, then the downstream server
246 notices it immediately and mirrors the "1h" file but then leaves the
247 rmirror loop and continues with "6h" on the next round through. This
248 behaviour goes on until it reaches the Z file. Because we eat the Z file
249 piecemeal we leave the loop after a while but after that everybody seems
250 to have forgotten that there is some work left to be done. Bad, bad,
251 bad.
253 ETOOCONFUSING better name?
255 Recentfile::get_remote_recentfile_as_tempfile OK
256 Recentfile::resolve_recentfilename split_rfilename DONE
258 Recent::_principal_recentfile_object _principal_recentfile_fromremote
259 Recent::principal_recentfile OK
260 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
262 some void, some not, some not void but called in void context.
264 2009-04-10 Andreas J. Koenig <andk@cpan.org>
266 * Bug? When the dirtymark changes then the second tier hosts quickly
267 reset their DONE state and restart mirroring. But then they mirror
268 potentially outdated index files with inconsistent dirtymark because the
269 first tier box may be a bit behind on the large recentfiles.
271 This means we need a brake when iterating over the recentfiles that
272 refuses to use a recentfile with the wrong dirtymark.
274 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
276 * install the bin/* files and talk about rrr-overview somewhere.
278 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
279 created outside the target tree.
281 * Barbie asks for better logging capabilities
283 2009-03-29 Andreas J. Koenig <andk@cpan.org>
285 * known bug: cannot configure the Recentfile objects to keep delete
286 events. How to call the accessor for that? keep_delete_objects_forever?
288 Sounds acceptable. FIXED (but untested).
290 * known bug: keeps temporary index files lying around.
292 2009-03-24 Andreas J. Koenig <andk@cpan.org>
294 * Some equivalent for
296 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
298 ? DONE
300 2009-03-22 Andreas J. Koenig <andk@cpan.org>
302 * k81 is again client of k75 with different parameters.
304 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
305 about it all the time and nobody reacts accordingly; (2) lots of temp
306 files get created and not removed; culprit the new call to
307 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
308 caller has to remove the tempfile after use.
310 Need the drawing board.
312 PARTLY FIXED: retracting the idea to call
313 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
314 and unseed such that uptodate now means "we have mirrored this
315 recentfile's bunch of files" and seeded means "somebody believes the
316 index file for this rf needs be refreshed". This seems to work now.
318 Still collecting temp files that nobody cares to depollute.
320 2009-03-21 Andreas J. Koenig <andk@cpan.org>
322 * Bug?: should it be harder than it is atm to set the timestamp to the
323 future? YES, FIXED
325 * bug with native integers:
327 -
328 epoch: 997872011
329 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
330 type: new
331 -
332 epoch: 1195248431
333 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
334 type: new
336 Native integer broke when native math was turned off. FIXED
338 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
339 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
340 in this area.
342 Yes, records out of order:
344 447003 -
345 447004 epoch: 995885533
346 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
347 447006 type: new
348 447007 -
349 447008 epoch: 995890358
350 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
351 447010 type: new
352 447011 -
353 447012 epoch: 995892221
354 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
355 447014 type: new
357 FIXED with sanity check and later with the integer fix.
359 * Bug: want the index files in a .recent directory
361 * Bug: lots of dot files are not deleted in time
363 * possible test case: can a delete change the timestamp? This would
364 probably break the order of events.
366 2009-03-20 Andreas J. Koenig <andk@cpan.org>
368 * 1233701831.34486 what's so special about this number/string? It
369 belongs to
370 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
371 atm lives in Y,Q, and Z.
373 It is the first entry after id/--skip-locking which has timestamp
374 1234164228.11325 which represents a file that doesn't exist anymore.
376 * another bug:
378 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
379 _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
380 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
381 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
382 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
383 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
384 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
385 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
386 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
387 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
388 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
389 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
390 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
392 FIXED, it was the "--skip-locking" file where manual intervention was participating
394 * bug on the mirroring slave: when the dirtymark gets increased we
395 probably do not reset the done intervals. The mirrorer stays within
396 tight bounds where it tries to sync with upstream and never seems to
397 finish. In the debugging state file I see lots of identical intervals
398 that do not get collapsed. When I restart the mirrorer it dies with:
400 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
401 _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
402 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
403 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
404 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
405 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
406 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
407 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
408 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
409 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
411 and debugging stands at
413 At: before
414 Brfinterval: Z
415 Ci: 227
416 Dre+1:
417 epoch: 1237400802.5789
418 path: id/J/JO/JOHND/CHECKSUMS
419 type: new
420 Dre-0:
421 epoch: 1237400807.97514
422 path: id/M/MI/MIYAGAWA/CHECKSUMS
423 type: new
424 Dre-1:
425 epoch: 1237400817.94363
426 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
427 type: new
428 Eintervals:
429 -
430 - 900644040
431 - 900644040
432 - []
434 and it is reproducable.
436 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
437 pause has a fresh one.
439 FIXED, it was the third anded term in each of the ifs in the IV block in
440 _register_one: with that we make sure that we do not stamp on valuable
441 interval data.
443 2009-03-17 Andreas J. Koenig <andk@cpan.org>
445 * done: verified the existence of the floating point bug in bleadperl
446 and verified that switching from YAML::Syck to YAML::XS does not resolve
447 it.
449 BTW, the switch was doable with
451 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
453 and should be considered as a separate TODO
455 * todo: integrate a dirty update with two aggregate calls before
456 unlocking for frictionless dirtying DONE
458 * todo: start the second rsync daemon on pause DONE
460 * todo: move index files to .recent: this cannot simply be done by
461 setting filenameroot to .recent/RECENT. Other parts of the modules rely
462 on the fact that dirname(recentfile) is the root of the mirrored tree.
464 2009-03-16 Andreas J. Koenig <andk@cpan.org>
466 * What was the resolution of the mirror.pl delete hook bug? Do we call
467 the delete hook when pause removes a file from MUIR?
469 * Today on pause: Updating 2a13fba..29f284d and installing it for
470 /usr/local/perl-5.10.0{,-RC2}
472 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
473 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
475 2009-03-15 Andreas J. Koenig <andk@cpan.org>
477 * currently recent_events has the side effect of setting dirtymark
478 because it forces a read on the file. That should be transparent, so
479 that the dirtymark call always forces a cache-able(?) read.
481 * The bug below is -- after a lot of trying -- not reproducible on a
482 small script, only in the large test script. The closest to the output
483 below was:
485 #!perl
486 use strict;
487 use Devel::Peek;
488 my $x = "01237123229.8814";
489 my($l,$r);
490 for ($l,$r) {
491 $_ = "x"x34;
492 }
493 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
494 $r = int $r;
495 $l = "1237123231.22458";
496 $r = "1237123231.22458";
497 1 if $l/1.1;
498 Devel::Peek::Dump $l;
499 Devel::Peek::Dump $r;
500 Devel::Peek::Dump $x = $l <=> $r;
501 die "BROKE" if $x;
502 __END__
504 The checked in state at c404a85 fails the test with my
505 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
506 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
507 in the test script always consistent.
509 * Todo: write a test that inserts a second dirty file with an already
510 existing timestamp. DONE
512 * Bug in perl 5.10 on my 64bit box:
514 DB<98> Devel::Peek::Dump $l
515 SV = PVMG(0x19e0450) at 0x142a550
516 REFCNT = 2
517 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
518 IV = 0
519 NV = 1237123231.22458
520 PV = 0x194ce70 "1237123231.22458"\0
521 CUR = 16
522 LEN = 40
524 DB<99> Devel::Peek::Dump $r
525 SV = PVMG(0x19e0240) at 0x142a3e8
526 REFCNT = 2
527 FLAGS = (PADMY,POK,pPOK)
528 IV = 1237123229
529 NV = 1237123229.8814
530 PV = 0x19ff900 "1237123231.22458"\0
531 CUR = 16
532 LEN = 40
534 DB<100> Devel::Peek::Dump $l <=> $r
535 SV = IV(0x19ea6e8) at 0x19ea6f0
536 REFCNT = 1
537 FLAGS = (PADTMP,IOK,pIOK)
538 IV = -1
540 DB<101> Devel::Peek::Dump $l
541 SV = PVMG(0x19e0450) at 0x142a550
542 REFCNT = 2
543 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
544 IV = 0
545 NV = 1237123231.22458
546 PV = 0x194ce70 "1237123231.22458"\0
547 CUR = 16
548 LEN = 40
550 DB<102> Devel::Peek::Dump $r
551 SV = PVMG(0x19e0240) at 0x142a3e8
552 REFCNT = 2
553 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
554 IV = 1237123231
555 NV = 1237123231.22458
556 PV = 0x19ff900 "1237123231.22458"\0
557 CUR = 16
558 LEN = 40
560 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
562 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
564 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
566 * $obj->merge ($other) needs to learn about equal epoch which may happen
567 since dirty_epoch intruded.
569 * Wontfix anytime soon: I think we currently do not support mkdir. Only
570 files!
572 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
574 * Todo: continue working on update(...,$dirty_epoch). It must be
575 followed by a fast_aggregate! DONE
577 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
579 * maybe we need a closest_entry or fitting_interval or something like
580 that. We want to merge an event into the middle of some recentfile.
581 First we do not know which file, then we do not know where to lock,
582 where to enter the new item, when and where to correct the dirtymark.
584 So my thought is we should first find which file.
586 Another part of my brain answers: what would happen if we would enter
587 the new file into the smallest file just like an ordinary new event,
588 just as an old event?
590 (1) we would write a duplicate timestamp? No, this would be easy to
591 avoid
593 (2) we would make the file large quickly? Yes, but so what? We are
594 changing the dirtymark, so are willing to disturb the downstream hosts.
596 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
598 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
600 leak!
602 https://rt.cpan.org/Ticket/Display.html?id=41199
604 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
606 Strangest output being HAKANARDO who managed to upload
608 Here is a better oneliner that includes also the first line of each
609 finding:
611 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
613 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
615 * memory leak in the syncher? It currently weighs 100M.
617 Update 2008-11-02:
619 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
622 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
624 * lookup by epoch and by path and use this ability on the pause to never
625 again register a file twice that doesn't need it. Let's call it
626 contains().
628 * after the dirtymark is done: fill up recentfiles with fake (historic)
629 entries; fill up with individual corrections; algorithm maybe to be done
630 with bigfloat so that we can always place something in the middle
631 between two entries. Before we must switch to bigfloat we could try to
632 use Data::Float::nextup to get the.
634 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
635 directly.
637 * dirtymark now lives in Recentfile, needs to be used in rmirror.
639 * find out why the downloader died after a couple of hours without a net
640 connection. Write a test that survives the not-existence of the other
641 end forever.
643 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
645 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
646 it makes less sense with tight coupling for secondary files. But asking
647 the server what the current epoch is might be cheaper on HTTP than on
648 rsync. (Needs to be evaluated)
650 * remove the 0.00 from the verbose overview in the Merged column in the
651 Z row. DONE
653 * write tests that expose the problems of the last few days: cascading
654 client/server roles, tight coupling for secondary RFs, deletes after
655 copies.
657 * Some day we might want to have policy options for the slave:
658 tight/loose/no coupling with upstream for secondary RFs. tight is what
659 we have now. loose would wait until a gap occurs that can be closed.
661 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
663 * revisit all $rfs->[$i+1] places if they now make sense still
665 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
667 * another bug is the fact that the mirror command deletes files before
668 it unhides the index file, thus confusing downstream slaves. We must not
669 delete before unhiding and must delete after unhiding. FIXED.
671 * new complication about the slave that is playing a server role.
672 Currently we mirror from newest to oldest with a hidden temporary file
673 as index. And when one file is finished, we unhide the index file.
674 Imagine the cascading server/slave is dead for a day. It then starts
675 mirroring again with the freshest thing and unhides the freshest index
676 file when it has worked through it. In that moment it exposes a time
677 hole. Because it now works on the second recentfile which is still
678 hidden.
680 We currently do nothing special to converge after such a drop out. At
681 least not intentionally and robustly and thought through.
683 The algorithm we use to seed the next file needs quite a lot of more
684 robustness than it currently has. Something to do with looking at the
685 merged element of the next rf and when it has dropped off, we seed
686 immediately. And if it ramains dropped off, we seed again, of course.
688 Nope, looking from smaller to larger RFS we look at the merged element
689 of this RF and at the minmax/max element of the next RF. If that
690 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
691 seeding it.
693 And we need a public accessor seed and unseed or seeded. But not the mix
694 of public and private stuff that then is used behind the back.
696 And then the secondary* stuff must go.
698 And we must understand what the impact is on the DONE system. Can it go
699 unnoticed that there was a hole? And could the DONE system have decided
700 the hole is covered? This should be testable with three directories where
701 the middle stops working for a while. Done->merge is suspicious, we must
702 stop it from merging non-conflatable neighbors due to broken continuity.
704 FIXED
706 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
708 * Slaven suggests to have the current epoch or the whole current
709 recentfile available from the HTTP server and take it away with
710 keepalive. This direction goes the granularity down to subseconds.
712 We might want to rewrite everything to factor out transport and allow
713 the whole thing to run via HTTP.
715 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
717 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
718 symlink?
720 * smoker on k81 fetching from k75 to verify cascading works. See
721 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
723 * maybe the loop should wait for CHECKSUMS file after every upload. And
724 CPAN.pm needs to deal with timestamps in the future.
726 * do not forget the dirtymark!
728 Text: have a new flag on recentfiles with the meaning: if this
729 changes, you're required to run a full rsync over all the files. The
730 reason why we set it would probably be: some foul happened. we injected
731 files in arbitrary places or didn't inject them although they changed.
732 The content of the flag? Timestamp? The relation between the
733 recentfiles would have to be inheritance from the principal, because any
734 out of band changes would soon later propagate to the next recentfile.
736 By upping the flag often one can easily ruin the slaves.
738 last out of band change? dirtymark?
740 Anyway, this implies that we read a potentially existing recentfile
741 before we write one.
743 And it implies that we have an eventloop that keeps us busy in 2-3
744 cycles, one for current stuff (tight loop) and one for the recentfiles
745 (cascade when principal has changed), one for the old stuff after a
746 dirtymark change.
748 And it implies that the out-of-band change in any of the recentfiles
749 must have a lock on the principal file and there is the place to set the
750 dirtymark.
752 * start a FAQ, especially quick start guide questions. Also to aid those
753 problematic areas where we have no good solution, like the "links"
754 option to rsync.
756 * wish feedback when we are slow.
758 * reduce mccabe
760 * Remove a few DEBUG statements.
762 * The multiple-rrr way of doing things needs a new option to rmirror,
763 like piecemeal or so. Not urgent because after the first pass through,
764 things run smoothely. It's only ugly during the first pass.
766 * I have the suspicion that the code is broken that decides if the
767 neighboring RF needs to be seeded. I fear when too much time has gone
768 between two calls (in our case more than one hour), it would not seed
769 the neighbor. Of course this will never be noticed, so we need a good
770 test for it.
772 * local/localroot confusion: I currently pass both options but one must
773 do.
775 * accounts for early birds on PAUSE rsync daemon.
777 * hardcoded 20 seconds
779 * who mirrors the index? DOING now.
781 * which CPAN mirrors offer rsync?
783 * visit all XXX, visit all _float places
785 * rename the pathdb stuff, it's too confusing. No idea how.
787 * rrr-inotify, backpan, rrr-register
789 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
791 * current bugs: the pathdb seems to get no reset, the seeding of the
792 secondaryttl stuff seems not to have an effect. Have helped myself with
793 a rand(10), need to fix this back. So not checked in. Does the rand
794 thing even help?
796 The rand thing helps. The secondaryttl stuff was in the wrong line,
797 fixed now.
799 The pathdb stuff was because I called either _pathdb or __pathdb on the
800 wrong object. FIXED now.
802 * It's not so beautiful if we never fetch the recentfiles that are not
803 the principal, even if this is correct behaviour. We really do not need
804 them after we have fetched the whole content.
806 OK, we want a switch for that: secondaryttl DONE
808 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
810 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
811 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
812 it lets 40 follow. FIXED
814 * See that the long running process really only updates the principal
815 file unless it has missed a timespan during which something happened. If
816 nothing happened, it must notice even when it misses the timespan. DONE
818 * we must throw away the pathdb when we have reached the end of Z. From
819 that moment we can have a very small pathdb because the only reason for
820 a pathdb is that we know to ignore old records in old files. We won't
821 need this pathdb again before the next full pass over the data is
822 necessary and then we will rebuild it as we go along. DONE
824 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
826 * I think, Done::register_one is doing wrong in that it does not
827 conflate neighboring pieces. The covered() method cannot do this because
828 it has no recent_events array at hand. But register_one has it and could
829 do it and for some reason misses to do it (sometimes).
831 This means that the three tests I just wrote can probably not survive
832 because they test with an already broken Done structure.
834 The art now is to detect how it happens, then to reproduce, then write a
835 test, then fix it.
837 So from the logfile this is what happens: we have a good interval with
838 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
839 top of it. Locally we now mirror F2 and open a new done interval for it.
840 Then we mirror F1 but this time with the timestamp T1b. And when we then
841 try to close the gap, we do not find T1 but instead something older. We
842 should gladly accept this older piece and this would fix this bug.
844 FIXED
846 * bug to fix: when the 1h file changes while rmirror is running, we do
847 correctly sync the new files but never switch to the 6h file but rather
848 stay in a rather quick loop that fetches the 1h file again and again.
850 Is it possible that we initialize a new object? Or does
851 get_remote_recentfile_as_tempfile overwrite something in myself?
853 Want a new option: _runstatusfile => $file which frequently dumps the
854 state of all recentfiles to a file.
856 FIXED
858 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
860 * Todo: now teach update to verify the timestamp is about to write
861 against the previous and use _increase_a_bit if it doesn't comply with
862 strict monotony. DONE
864 * The problem of rounding. So far perl's default precision was
865 sufficient. One day it won't be. FakeFloat has an easy job when it is
866 only reading and other machines have written correctly. But when we want
867 to write a floating point number that is a bit larger than the other
868 one, then we need our own idea of precision.
870 Slaven said: just append a "1". This might be going towards the end of
871 usability too quickly. I'd like something that actually uses the decimal
872 system. Well, appending a 1 also does this but...
874 E.g. we have 1.0. nextup on this architecture is starting with
875 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
876 taken the 1.0000000000000003 and the next user comes and the time tells
877 him 1.0 again. He has to beat my number without stepping over the
878 nextup. This is much less space than I had when I chose 1,2,3.
880 What is also irritating is that nextup is architecture dependent. The
881 128 bit guy must choose very long numbers to fit in between whereas the
882 other one with 16 bit uses larger steps. But then the algorithm is the
883 same for both, so that would be a nice thing.
885 I see two situation where we need this. One is when Time::HiRes returns
886 us a value that is <= the last entry in our recentfile. In this case
887 (let's call it the end-case) we must fill the region between that number
888 and the next higher native floating point number. The other is when we
889 inject an old file into an old recentfile (we would then also set a new
890 dirtymark). We find the integer value already taken and need a slightly
891 different one (let's call it the middle-case). The difference between
892 the two situations is that the next user will want to find something
893 higher than my number in the end-case and something lower than my number
894 in the middle case.
896 So I suggest we give the function both a value and an upper bound and it
897 calculates us a primitive middle. The upper bound in the middle-case is
898 the next integer. The upper bound on the end-case is the nextup floating
899 point number. But the latter poses another problem: if we have occupied
900 the middle m between x and nextup(x), then the nextup(m) will probably
901 not be the same as nextup(x) because some rounding will take place
902 before the nextup is calculated and when the rounding reaches the
903 nextup(x), we will end up at nextup(nextup(x)).
905 So we really need to consider the nextup and the nextdown from there and
906 then the middle and that's the number we may approach asymptotically.
907 Ugly. But DONE.
909 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
911 * consider deprecating the use of RECENT.recent as a symlink. It turns
912 out to need extra hoops with the rsync options and just isn't worth it.
913 Or maybe these extra hoops are needed anyway for the rest of the tree?
914 Nope, can't be the case because not all filesystems support symlinks.
916 But before doing the large step, I'll deprecate the call of
917 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
918 only introduced to resolve RECENT.recent and complicates the routine far
919 beyond what it deserves.
921 DONE. Won't deprecate RECENT.recent, just moved its handling to the
922 supervisor.
924 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
926 * I think it's a bug that the rsync_option links must be set to true in
927 order to support RECENT.recent and that nobody cares to set it
928 automatically. Similar for ignore_link_stat_errors.
930 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
932 * Todo: find all todos together and make a plan what is missing for a
933 release.
935 - verifytree or something like that. fsck maybe.
937 - rersyncrecent, the script itself? What it do?
939 - a way to only mirror the recentfiles without mirroring the whole
940 remote system such that people can decide to mirror only partially see
941 also 2008-08-30. .shadow-xxx directory? this also needed for a
942 filesystem that is still incomplete and might need the mirrorfiles for
943 lookup(?)
945 - long living objects that mirror again and again. Inject something
946 into ta, see how it goes over to tb.
948 - how do we continue filling up the DONE system when we use an object
949 for the second time? "fully covered" and "uptodate" or new terminology.
951 - overview called on the wrong file should be understandable
953 - the meta data field that must change when we fake something up so that
954 the downstream people know they have to re-fetch everything.
956 - how tolerant are we against missing files upstream? how do we keep
957 track? there are legitimate cases where we did read upstream index right
958 before a file got deleted there and then find that file as new and want
959 it. There are other cases that are not self healing and must be tracked
960 and bugreported.
962 - how, exactly, do we have to deal with deletes? With rsync errors?
964 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
965 authors) failed: No such file or directory (2)
967 The file above is a delete in 1h and a new in file 1M and the
968 delete in the locally running rmirror did not get propagated to the 1M
969 object. Bug. And the consequence is a standstill.
971 It seems that a slave that works with a file below the principal needs
972 to merge things all the way up to get rid of later deletes. Or keep
973 track of all deletes and skip them later. So we need a trackdeletes.pm
974 similar to the done.pm?
976 see also 2008-08-20 about spurious deletes that really have no add
977 counterpart and yet they are not wrong.
979 - consider the effect when resyncing the recentfile takes longer than
980 the time per loop. Then we never rsync any file. We need to diagnose
981 that and force an increase of that loop time. But when we later are fast
982 enough again because the net has recovered, then we need to switch back
983 to original parameters. ERm, no, it's enough to keep syncing at least
984 one file before refetching an index file.
986 - remember to verify that no temp files are left lying around and the
987 signal handler
989 - status file for not long running jobs that want to track upstream with
990 a, say, cronjob.
992 - revisit all XXX _float areas and study Sub::Exporter DONE
994 - persistent DB even though we just said we do not need it. Just for
995 extended capabilities and time savings when, for example, upstream
996 announces a reset and we get new recentfiles and could then limit
997 ourselves to a subset of files (those that have a changed epoch) in a
998 first pass and would only then do the loop to verify the rest. Or
999 something.
1001 * Todo: aggregate files should know their feed and finding the principal
1002 should be done stepwise. (?)
1004 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
1005 left locks around. DONE
1007 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1009 * maybe extend the _overview so that it always says if and where the
1010 last file is in the next file and where the next event in the next rf
1011 would lie. No, don't like this anymore. REJECT
1013 * take the two new redundant tests out again, only the third must
1014 survive. DONE
1016 * Todo: add a sanity check if the merged structure is really pointing to
1017 a different rf and that this different rf is larger. DONE
1019 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1021 * now test, if they are overlapping. And test if there is a file in the
1022 next rf that would fit into this rf's interval.
1024 1h 1222324012.8474 1222322541.7963 0.4086
1025 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
1026 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
1027 1W 1222313218.3626 1221708477.5829 167.9835
1029 I suspect that somebody writes a merged timestamp without having merged
1030 and then somebody else relies on it.
1032 If aggregate is running, the intervals must not be extravagated, if it
1033 is not running, there must not be bounds, the total number of events in
1034 the system must be counted and must be controlled throughout the tests.
1035 That the test required the additional update was probably nonsense,
1036 because aggregate can cut pieces too. FIXED & DONE
1038 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1040 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1041 changed. FIXED
1043 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1045 * Most apparent bug at the moment is that the recentfiles are fetched
1046 too often. Only the principal should be fetched and if it has not
1047 changed, the others should not be refetched. ATM I must admit that I'm
1048 happy that we refetch more often than needed because I can more easily
1049 fix bugs while the thing is running.
1051 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1052 done but the done system does not know about it. The reason for the
1053 failure is not known and we never reach the status uptodate because of
1054 this. We must get over it.
1056 Later it turns out that the origin server had a bug somewhere.
1057 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1058 array of the recentfile one position above. The bug was my own.
1060 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1062 * There is the race condition where the server does a delete and the
1063 slave does not yet know and then tries to download it because he sees
1064 the new. So for this time window we must be more tolerant against
1065 failure. If we cannot download a file, we should just skip it and should
1066 not retry immediately. The whole system should discover the lost thing
1067 later. Keeping track with the DONE system should really be a no brainer.
1069 But there is something more: the whole filesystem is a database and the
1070 recentfiles are one possible representation of it. It's a pretty useful
1071 representation I think that's why I have implemented something around
1072 it. But for strictly local operation it has little value. For local
1073 operation we would much rather have a database. So we would enter every
1074 recentfile reading and every rsync operation and for every file the last
1075 state change and what it leads to. Then we would always ignore older
1076 records without the efforts involved with recentfiles.
1078 The database would have: path,recentepoch,rsyncedon,deletedon
1080 Oh well, not yet clear where this leads to.
1082 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1084 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1085 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1087 * statusfile, maybe only on demand, alone to have a sharp debugging
1088 tool. It is locked and all recentfiles dump themselves into it and we
1089 can build a viewer that lets us know where we stand and what's inside.
1091 * remember: only the principal recentfile needs expiration, all others
1092 shall be expired by principal if it discovers that something has move
1093 upstream.
1095 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1097 * Always check if we stringify to a higher value than in the entry
1098 before. DONE
1100 * And in covered make an additional check if we would be able to see a
1101 numerical difference between the two numbers and if we can't then switch
1102 to a different, more expensive algorithm. Do not want to be caught by
1103 floating surprises. DONE
1105 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1107 * caching has several aspects here: we can cache the interval of the
1108 recentfile which only will change when the mtime of the file changes. We
1109 must re-mirror the recentfile when its ttl has expired. Does have_read
1110 tell you anything? It counts nothing at all. Only the mtime is
1111 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1112 thing the have_mirrored because it is unrelated to the mtime.
1114 * Robustness of floating point calculations! I always thought that the
1115 string calculated by the origin server for the floating representation
1116 of the epoch time is just a string. When we convert it to a number and
1117 later back to a string, the other computer might come to a different
1118 conclusion. This must not happen, we want to preserve it under any
1119 circumstances. I will have to write tests with overlong sequences that
1120 get lost in arithmetic and must see if all still works well. DONE
1122 But one fragile point remains: if one host considers a>b and the other
1123 one considers them == but no eq. To prevent this, we must probably do
1124 some extra homework. DONE
1126 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1128 * the concept of tracking DONE needs an object per recentfile that has
1129 something like these methods:
1131 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1133 covered() register() covered()
1135 The unclear thing is how we translate points in time into intervals. We
1136 could pass a reference to the current recent_events array when running
1137 we_have(xxx) and let the DONE object iterate over it such that it only
1138 has to store a list of intervals that can melt into each other. Ah, even
1139 passing the list together with a list of indexes seems feasiable.
1141 Or maybe ask for the inverted list?
1143 Whenever the complete array is covered by the interval we say we are
1144 fully covered and if the recentfile is not expired, we are uptodate.
1146 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1148 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1150 * need a way to "return" the next entry after the end of a list. When
1151 the caller says "before" or "after" we would like to know if he could
1152 cover that interval/threshold or not because this influences the effect
1153 of a newer timestamp of that recentfile. DONE with $opt{info}.
1155 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1157 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1159 loop implies tixu (?). Nope, something like --statefile decides. Per
1160 default we do ...?
1162 T test, I init, X including removals, U nodirtymark
1164 So we have no concept of dirtymarks, we only trust that since we are
1165 running we have observed everything steadily. But people will not let
1166 this program run forever so we must consider both startup penalty and
1167 book keeping for later runs. We keep this for later. For now we write a
1168 long running mirror that merges several intervals.
1170 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1172 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1173 the intervals!
1175 * rersyncrecent, the script: default to one week. The name of the switch
1176 is --after. Other switches? --loop!
1178 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1180 * need a switch --skip-deletes (?)
1182 * need a switch --enduser that tells us that the whole tempfile
1183 discipline is not needed when there is no downstream user. (?)
1185 Without this switch we cannot have a reasonable recent.pl that just
1186 displays the recent additions. Either we accept to download everything.
1187 Or we download temporary files without the typical rsync protocol
1188 advantages.
1190 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1191 File::Temp::tempdir, this might be a win.
1193 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1195 * apropos missing: we have no push, we never know the downstream
1196 servers. People who know their downstream hosts and want to ascertain
1197 something will want additional methods we have never thought about, like
1198 update or delete a certain file.
1200 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1202 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1203 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1204 difference.
1206 * rsync options: it's a bit of a pain that we usually need several rsync
1207 options, like compress, links, times, checksum and that there is no
1208 reasonable default except the original rsync default. I think wee can
1209 safely assume that the rsync options are shared between all recentfile
1210 instances within one recent tree.
1212 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1214 * deletes: if a delete follows an add quickly enough it may happen that
1215 a downstream mirror did not see the add at all! It seems this needs to
1216 be mentioned somewhere. The point here is that even if the downstream is
1217 never missing the principal timeframe it may encounter a "delete" that
1218 has no complimentary "add" anywhere.
1220 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1222 * I suspect the treat of metadata is incorrect during read or something.
1223 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1224 contained more than 6 hours worth of data. At 06:08 we merged into the
1225 1d file. We need to take snapshots of the 6h file over the course of an
1226 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1227 enough.
1229 Much worse: watching the 1h file: right at the moment (at 06:35) it
1230 covers 1218867584-1219120397 which is 70 hours.
1232 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1233 08:19:44 2008, that is when I checked out last time, so it seems to be
1234 aggregating and never truncating?
1236 No, correct is: it is never truncating; but wrong is: it is aggregating.
1237 It does receive a lot of events from time to time from a larger file.
1238 Somehow a large file gets merged into the small one and because the
1239 "meta/merged" attribute is missing, nobody is paying attention. I
1240 believe that I can fix this by making sure that metadata are honoured
1241 during read. DONE and test adjusted.
1243 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1245 * grand renaming plan
1247 remotebase => remoteroot to fit well with localroot DONE
1248 local_path() => localroot seems to me should already work DONE
1249 recentfile_basename => rfilename no need to stress it has no slash DONE
1251 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1252 kern? It would anyway need a deprecation cycle because it is an
1253 important constructor.
1255 * I like the portability that Data::Serializer brings us but the price
1256 is that some day we might find out that it is slowing us a bit. We'll
1257 see.
1259 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1261 * should we not enter the interval of the principal (or the interval of
1262 the merging file?) in every aggregated/merged file?
1264 * we should aim at a first release and give up on thinking about
1265 sanitizing stuff and zloop. Let's just admit that a full traditional
1266 rsync is the only available sanitizer ATM. Otherwise it's complicated
1267 stuff: sanitizing on the origin server, sanitizing on the slaves,
1268 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1269 get the basics out before this becomes a major cause for mess.
1271 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1273 * On OSes not supporting symlinks we expect that RECENT.recent contains
1274 the contents of the principal recentfile. Actually this is identical on
1275 systems supporting symlinks. Simple, what follows from that is that we
1276 need to keep the serializer in the metadata because we cannot read it
1277 from the filename, doesn't it? Of course not. It's a chicken and egg
1278 problem. This leaves us with the problem to actually parse the
1279 serialized data to find out in which format it is. So who can do the 4
1280 or 5 magics we wanted to support? File::LibMagic?
1282 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1284 * remotebase and recentfile_basename are ugly names. Now that we need a
1285 word for the shortest/principal/driving recentfile too we should do
1286 something about it.
1288 localroot is good. rfile is good. local_path() is bad, local_path($path)
1289 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1290 already deprecated.
1292 Up to now remotebase was the string that described the remote root
1293 directory in rsync notation, like pause.perl.org::authors. And
1294 recentfile_basename was "RECENT-1h.yaml".
1296 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1298 * The test that was added in today's checkin is a good start for a test
1299 of rmirror. We should have more methods in Recent.pm: verify,
1300 addmissingfiles. We should verify the current tree, then rmirror it and
1301 then verifytree the copy. We could then add some arbitrary file and let
1302 it be discovered by addmissingfiles, then rmirror again and then
1303 verifytree the copy again.
1305 Then we could start stealing from csync2 sqlite database [no port to
1306 OSX!] and fill a local DB. And methods to compare the database with the
1307 recentfiles. Our strength is that in principle we could maintain state
1308 with a single float. We have synced up to 1234567890.123456. If the Z
1309 file does not add new files all we have to do is mirror the new ones and
1310 delete the goners.
1312 This makes it clear that we should extend current protocol and declare
1313 that we cheat when we add files too late, just to help the other end
1314 keeping track. Ah yes, that's what was meant when zloop was mentioned
1315 earlier.
1317 Maybe need to revisit File::Mirror to help me with this task.
1319 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1321 * There must be an allow-me-to-truncate flag in every recentfile.
1322 Without it one could construct a sequence of updates winning the locking
1323 battle against the aggregator. Only if an aggregator has managed to
1324 merge data over to the next level, truncating can be allowed. DONE with
1325 accessor merged.
1327 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1329 * We should probably guarantee that no duplicates enter the aggregator
1330 array.
1332 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1334 * To get merge operation faster would need a good benchmark test. What
1335 02 spits out isn't reliable enough and is dominated by many other
1336 things. Between
1338 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1339 Author: Andreas J. Koenig <k@k75.(none)>
1340 Date: Sat Aug 2 07:58:04 2008 +0200
1342 and
1344 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1345 Author: Andreas J. Koenig <k@k75.(none)>
1346 Date: Sat Aug 2 11:40:29 2008 +0200
1348 there was a lot of trying but the effect is hardly measurable with
1349 current tests.
1351 * overhead of connecting seems high. When setting
1352 max_files_per_connection to 1 we see that.
1354 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1356 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1358 25138.2215600014/3600 = 6.98283932222261
1360 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1361 be a bug in the aggregator. FIXED
1363 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1365 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1366 we have it, can we mirror it, mirror it!
1368 I fear this needs a new class which might be called
1369 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1370 and treat them as an entity. I realize that a single recentfile may be
1371 sufficient for certain tasks and that it is handy for the low level
1372 programmer but it is not nice to use. If there is a delete in the 1h
1373 file then the 6h file still contains it. Seekers of the best information
1374 need to combine at least some of the recentfiles most of the time.
1376 There is the place for the Z loop!
1378 But the combination is something to collect in a database, isn't it. Did
1379 csync2 just harrumph?
1381 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1383 * it just occurred to me that hosts in the same mirroring pool could
1384 help out each other even without rewriting the recentfile. Just fetch
1385 the stuff to mirror from several places, bingo. But that's something
1386 that should rather live in a separate package or in rsync directly.
1388 * cronjobs are unsuited because with ntp they would all come at the full
1389 minute and disturb each other. Besides that I'd hate to have a backbone
1390 with more than a few seconds latency.
1392 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1394 * a second rsync server with access control for PAUSE. Port? 873 is the
1395 standard port, let's take 8873.
1397 * if there were a filesystem based on this, it would have a slow access
1398 to inexistent files. It would probably provide wrong readdir (only based
1399 on current content) or also a slow one (based on a recentfile written
1400 after the call). But it would provide fast access to existing files. Or
1401 one would deliberately allow slightly blurred answers based on some
1402 sqlite reflection of the recentfiles.
1404 * todo: write a variant of mirror() that combines two or more
1405 recentfiles and treats them like one
1407 * todo: signal handler to remove the tempfile
1409 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1411 * now that we have the symlink I forgot how it should be used in
1412 practice.
1414 * the z loop: add missing files to Z file. Just append them (instead of
1415 prepending). So one guy prepends something from the Y file from time to
1416 time and another guy appends something rather frequently. Collecting
1417 pond. When Y merges into Z, things get epoch and the collecting pond
1418 gets smaller. What exactly are "missing files"?
1420 take note of current epoch of the alpha file, let's call it the
1421 recent-ts
1423 find all files on disk
1425 remove all files registered in the recentworld up to recent-ts
1427 remove all files that have been deleted after recent-ts according to
1428 recentworld
1430 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1432 * rersyncrecent might be a cronjob with a (locked) state file which
1433 contains things like after and maybe last z sync or such?
1435 rrr-mirror might be an alternative name but how would we justify the
1436 three Rs when there is no Re-Rsync-Recent?
1438 With the --loop parameter it is an endless loop, without it is no loop.
1439 At least this is simple.
1441 * todo: new accssor z-interval specifies how often the Z file is updated
1442 against the filesystem. We probably want no epoch stamp on these
1443 entries. And we want to be able to filter the entries (e.g. no
1444 by-modules and by-category tree)
1446 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1448 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1449 into Z. We do not need the other files filled up with filesystem
1450 contents though.
1452 * need interface to query for a file in order to NOT call update on
1453 PAUSE a second time within a short time.
1455 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1457 * recommended update interval? Makes no sense, is different for
1458 different users.
1460 * Moosify
1462 Local Variables:
1463 mode: change-log
1464 change-log-default-name: "Todo"
1465 tab-width: 2
1466 left-margin: 2
1467 End: