1 2009-03-22 Andreas J. Koenig <andk@cpan.org>
3 * k81 is again client of k75 with different parameters.
5 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
6 about it all the time and nobody reacts accordingly; (2) lots of temp
7 files get created and not removed; culprit the new call to
8 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
9 caller has to remove the tempfile after use.
11 Need the drawing board.
13 PARTLY FIXED: retracting the idea to call
14 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
15 and unseed such that uptodate now means "we have mirrored this
16 recentfile's bunch of files" and seeded means "somebody believes the
17 index file for this rf needs be refreshed". This seems to work now.
19 Still collecting temp files that nobody cares to depollute.
21 2009-03-21 Andreas J. Koenig <andk@cpan.org>
23 * Bug?: should it be harder than it is atm to set the timestamp to the
26 * bug with native integers:
30 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
34 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
37 Native integer broke when native math was turned off. FIXED
39 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
40 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
43 Yes, records out of order:
46 447004 epoch: 995885533
47 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
50 447008 epoch: 995890358
51 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
54 447012 epoch: 995892221
55 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
58 FIXED with sanity check and later with the integer fix.
60 * Bug: want the index files in a .recent directory
62 * Bug: lots of dot files are not deleted in time
64 * possible test case: can a delete change the timestamp? This would
65 probably break the order of events.
67 2009-03-20 Andreas J. Koenig <andk@cpan.org>
69 * 1233701831.34486 what's so special about this number/string? It
71 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
72 atm lives in Y,Q, and Z.
74 It is the first entry after id/--skip-locking which has timestamp
75 1234164228.11325 which represents a file that doesn't exist anymore.
79 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
80 _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
81 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
82 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
83 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
84 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
85 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
86 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
87 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
88 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
89 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
90 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
91 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
93 FIXED, it was the "--skip-locking" file where manual intervention was participating
95 * bug on the mirroring slave: when the dirtymark gets increased we
96 probably do not reset the done intervals. The mirrorer stays within
97 tight bounds where it tries to sync with upstream and never seems to
98 finish. In the debugging state file I see lots of identical intervals
99 that do not get collapsed. When I restart the mirrorer it dies with:
101 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
102 _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
103 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
104 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
105 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
106 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
107 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
108 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
109 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
110 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
112 and debugging stands at
118 epoch: 1237400802.5789
119 path: id/J/JO/JOHND/CHECKSUMS
122 epoch: 1237400807.97514
123 path: id/M/MI/MIYAGAWA/CHECKSUMS
126 epoch: 1237400817.94363
127 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
135 and it is reproducable.
137 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
138 pause has a fresh one.
140 FIXED, it was the third anded term in each of the ifs in the IV block in
141 _register_one: with that we make sure that we do not stamp on valuable
144 2009-03-17 Andreas J. Koenig <andk@cpan.org>
146 * done: verified the existence of the floating point bug in bleadperl
147 and verified that switching from YAML::Syck to YAML::XS does not resolve
150 BTW, the switch was doable with
152 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
154 and should be considered as a separate TODO
156 * todo: integrate a dirty update with two aggregate calls before
157 unlocking for frictionless dirtying
159 * todo: start the second rsync daemon on pause
161 * todo: move index files to .recent: this cannot simply be done by
162 setting filenameroot to .recent/RECENT. Other parts of the modules rely
163 on the fact that dirname(recentfile) is the root of the mirrored tree.
165 2009-03-16 Andreas J. Koenig <andk@cpan.org>
167 * What was the resolution of the mirror.pl delete hook bug? Do we call
168 the delete hook when pause removes a file from MUIR?
170 * Today on pause: Updating 2a13fba..29f284d and installing it for
171 /usr/local/perl-5.10.0{,-RC2}
173 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
174 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
176 2009-03-15 Andreas J. Koenig <andk@cpan.org>
178 * currently recent_events has the side effect of setting dirtymark
179 because it forces a read on the file. That should be transparent, so
180 that the dirtymark call always forces a cache-able(?) read.
182 * The bug below is -- after a lot of trying -- not reproducible on a
183 small script, only in the large test script. The closest to the output
189 my $x = "01237123229.8814";
194 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
196 $l = "1237123231.22458";
197 $r = "1237123231.22458";
199 Devel::Peek::Dump $l;
200 Devel::Peek::Dump $r;
201 Devel::Peek::Dump $x = $l <=> $r;
205 The checked in state at c404a85 fails the test with my
206 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
207 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
208 in the test script always consistent.
210 * Todo: write a test that inserts a second dirty file with an already
211 existing timestamp. DONE
213 * Bug in perl 5.10 on my 64bit box:
215 DB<98> Devel::Peek::Dump $l
216 SV = PVMG(0x19e0450) at 0x142a550
218 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
220 NV = 1237123231.22458
221 PV = 0x194ce70 "1237123231.22458"\0
225 DB<99> Devel::Peek::Dump $r
226 SV = PVMG(0x19e0240) at 0x142a3e8
228 FLAGS = (PADMY,POK,pPOK)
231 PV = 0x19ff900 "1237123231.22458"\0
235 DB<100> Devel::Peek::Dump $l <=> $r
236 SV = IV(0x19ea6e8) at 0x19ea6f0
238 FLAGS = (PADTMP,IOK,pIOK)
241 DB<101> Devel::Peek::Dump $l
242 SV = PVMG(0x19e0450) at 0x142a550
244 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
246 NV = 1237123231.22458
247 PV = 0x194ce70 "1237123231.22458"\0
251 DB<102> Devel::Peek::Dump $r
252 SV = PVMG(0x19e0240) at 0x142a3e8
254 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
256 NV = 1237123231.22458
257 PV = 0x19ff900 "1237123231.22458"\0
261 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
263 * Todo: reset "done" or "covered" and "minmax" after a dirty operation?
265 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
267 * $obj->merge ($other) needs to learn about equal epoch which may happen
268 since dirty_epoch intruded.
270 * Wontfix anytime soon: I think we currently do not support mkdir. Only
273 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
275 * Todo: continue working on update(...,$dirty_epoch). It must be
276 followed by a fast_aggregate!
278 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
280 * maybe we need a closest_entry or fitting_interval or something like
281 that. We want to merge an event into the middle of some recentfile.
282 First we do not know which file, then we do not know where to lock,
283 where to enter the new item, when and where to correct the dirtymark.
285 So my thought is we should first find which file.
287 Another part of my brain answers: what would happen if we would enter
288 the new file into the smallest file just like an ordinary new event,
289 just as an old event?
291 (1) we would write a duplicate timestamp? No, this would be easy to
294 (2) we would make the file large quickly? Yes, but so what? We are
295 changing the dirtymark, so are willing to disturb the downstream hosts.
297 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
299 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
303 https://rt.cpan.org/Ticket/Display.html?id=41199
305 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
307 Strangest output being HAKANARDO who managed to upload
309 Here is a better oneliner that includes also the first line of each
312 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
314 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
316 * memory leak in the syncher? It currently weighs 100M.
320 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
323 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
325 * lookup by epoch and by path and use this ability on the pause to never
326 again register a file twice that doesn't need it. Let's call it
329 * after the dirtymark is done: fill up recentfiles with fake (historic)
330 entries; fill up with individual corrections; algorithm maybe to be done
331 with bigfloat so that we can always place something in the middle
332 between two entries. Before we must switch to bigfloat we could try to
333 use Data::Float::nextup to get the.
335 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
338 * dirtymark now lives in Recentfile, needs to be used in rmirror.
340 * find out why the downloader died after a couple of hours without a net
341 connection. Write a test that survives the not-existence of the other
344 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
346 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
347 it makes less sense with tight coupling for secondary files. But asking
348 the server what the current epoch is might be cheaper on HTTP than on
349 rsync. (Needs to be evaluated)
351 * remove the 0.00 from the verbose overview in the Merged column in the
354 * write tests that expose the problems of the last few days: cascading
355 client/server roles, tight coupling for secondary RFs, deletes after
358 * Some day we might want to have policy options for the slave:
359 tight/loose/no coupling with upstream for secondary RFs. tight is what
360 we have now. loose would wait until a gap occurs that can be closed.
362 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
364 * revisit all $rfs->[$i+1] places if they now make sense still
366 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
368 * another bug is the fact that the mirror command deletes files before
369 it unhides the index file, thus confusing downstream slaves. We must not
370 delete before unhiding and must delete after unhiding. FIXED.
372 * new complication about the slave that is playing a server role.
373 Currently we mirror from newest to oldest with a hidden temporary file
374 as index. And when one file is finished, we unhide the index file.
375 Imagine the cascading server/slave is dead for a day. It then starts
376 mirroring again with the freshest thing and unhides the freshest index
377 file when it has worked through it. In that moment it exposes a time
378 hole. Because it now works on the second recentfile which is still
381 We currently do nothing special to converge after such a drop out. At
382 least not intentionally and robustly and thought through.
384 The algorithm we use to seed the next file needs quite a lot of more
385 robustness than it currently has. Something to do with looking at the
386 merged element of the next rf and when it has dropped off, we seed
387 immediately. And if it ramains dropped off, we seed again, of course.
389 Nope, looking from smaller to larger RFS we look at the merged element
390 of this RF and at the minmax/max element of the next RF. If that
391 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
394 And we need a public accessor seed and unseed or seeded. But not the mix
395 of public and private stuff that then is used behind the back.
397 And then the secondary* stuff must go.
399 And we must understand what the impact is on the DONE system. Can it go
400 unnoticed that there was a hole? And could the DONE system have decided
401 the hole is covered? This should be testable with three directories where
402 the middle stops working for a while. Done->merge is suspicious, we must
403 stop it from merging non-conflatable neighbors due to broken continuity.
407 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
409 * Slaven suggests to have the current epoch or the whole current
410 recentfile available from the HTTP server and take it away with
411 keepalive. This direction goes the granularity down to subseconds.
413 We might want to rewrite everything to factor out transport and allow
414 the whole thing to run via HTTP.
416 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
418 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
421 * smoker on k81 fetching from k75 to verify cascading works. See
422 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
424 * maybe the loop should wait for CHECKSUMS file after every upload. And
425 CPAN.pm needs to deal with timestamps in the future.
427 * do not forget the dirtymark!
429 Text: have a new flag on recentfiles with the meaning: if this
430 changes, you're required to run a full rsync over all the files. The
431 reason why we set it would probably be: some foul happened. we injected
432 files in arbitrary places or didn't inject them although they changed.
433 The content of the flag? Timestamp? The relation between the
434 recentfiles would have to be inheritance from the principal, because any
435 out of band changes would soon later propagate to the next recentfile.
437 By upping the flag often one can easily ruin the slaves.
439 last out of band change? dirtymark?
441 Anyway, this implies that we read a potentially existing recentfile
444 And it implies that we have an eventloop that keeps us busy in 2-3
445 cycles, one for current stuff (tight loop) and one for the recentfiles
446 (cascade when principal has changed), one for the old stuff after a
449 And it implies that the out-of-band change in any of the recentfiles
450 must have a lock on the principal file and there is the place to set the
453 * start a FAQ, especially quick start guide questions. Also to aid those
454 problematic areas where we have no good solution, like the "links"
457 * wish feedback when we are slow.
461 * Remove a few DEBUG statements.
463 * The multiple-rrr way of doing things needs a new option to rmirror,
464 like piecemeal or so. Not urgent because after the first pass through,
465 things run smoothely. It's only ugly during the first pass.
467 * I have the suspicion that the code is broken that decides if the
468 neighboring RF needs to be seeded. I fear when too much time has gone
469 between two calls (in our case more than one hour), it would not seed
470 the neighbor. Of course this will never be noticed, so we need a good
473 * local/localroot confusion: I currently pass both options but one must
476 * accounts for early birds on PAUSE rsync daemon.
478 * hardcoded 20 seconds
480 * who mirrors the index? DOING now.
482 * which CPAN mirrors offer rsync?
484 * visit all XXX, visit all _float places
486 * rename the pathdb stuff, it's too confusing. No idea how.
488 * rrr-inotify, backpan, rrr-register
490 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
492 * current bugs: the pathdb seems to get no reset, the seeding of the
493 secondaryttl stuff seems not to have an effect. Have helped myself with
494 a rand(10), need to fix this back. So not checked in. Does the rand
497 The rand thing helps. The secondaryttl stuff was in the wrong line,
500 The pathdb stuff was because I called either _pathdb or __pathdb on the
501 wrong object. FIXED now.
503 * It's not so beautiful if we never fetch the recentfiles that are not
504 the principal, even if this is correct behaviour. We really do not need
505 them after we have fetched the whole content.
507 OK, we want a switch for that: secondaryttl DONE
509 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
511 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
512 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
513 it lets 40 follow. FIXED
515 * See that the long running process really only updates the principal
516 file unless it has missed a timespan during which something happened. If
517 nothing happened, it must notice even when it misses the timespan. DONE
519 * we must throw away the pathdb when we have reached the end of Z. From
520 that moment we can have a very small pathdb because the only reason for
521 a pathdb is that we know to ignore old records in old files. We won't
522 need this pathdb again before the next full pass over the data is
523 necessary and then we will rebuild it as we go along. DONE
525 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
527 * I think, Done::register_one is doing wrong in that it does not
528 conflate neighboring pieces. The covered() method cannot do this because
529 it has no recent_events array at hand. But register_one has it and could
530 do it and for some reason misses to do it (sometimes).
532 This means that the three tests I just wrote can probably not survive
533 because they test with an already broken Done structure.
535 The art now is to detect how it happens, then to reproduce, then write a
538 So from the logfile this is what happens: we have a good interval with
539 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
540 top of it. Locally we now mirror F2 and open a new done interval for it.
541 Then we mirror F1 but this time with the timestamp T1b. And when we then
542 try to close the gap, we do not find T1 but instead something older. We
543 should gladly accept this older piece and this would fix this bug.
547 * bug to fix: when the 1h file changes while rmirror is running, we do
548 correctly sync the new files but never switch to the 6h file but rather
549 stay in a rather quick loop that fetches the 1h file again and again.
551 Is it possible that we initialize a new object? Or does
552 get_remote_recentfile_as_tempfile overwrite something in myself?
554 Want a new option: _runstatusfile => $file which frequently dumps the
555 state of all recentfiles to a file.
559 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
561 * Todo: now teach update to verify the timestamp is about to write
562 against the previous and use _increase_a_bit if it doesn't comply with
563 strict monotony. DONE
565 * The problem of rounding. So far perl's default precision was
566 sufficient. One day it won't be. FakeFloat has an easy job when it is
567 only reading and other machines have written correctly. But when we want
568 to write a floating point number that is a bit larger than the other
569 one, then we need our own idea of precision.
571 Slaven said: just append a "1". This might be going towards the end of
572 usability too quickly. I'd like something that actually uses the decimal
573 system. Well, appending a 1 also does this but...
575 E.g. we have 1.0. nextup on this architecture is starting with
576 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
577 taken the 1.0000000000000003 and the next user comes and the time tells
578 him 1.0 again. He has to beat my number without stepping over the
579 nextup. This is much less space than I had when I chose 1,2,3.
581 What is also irritating is that nextup is architecture dependent. The
582 128 bit guy must choose very long numbers to fit in between whereas the
583 other one with 16 bit uses larger steps. But then the algorithm is the
584 same for both, so that would be a nice thing.
586 I see two situation where we need this. One is when Time::HiRes returns
587 us a value that is <= the last entry in our recentfile. In this case
588 (let's call it the end-case) we must fill the region between that number
589 and the next higher native floating point number. The other is when we
590 inject an old file into an old recentfile (we would then also set a new
591 dirtymark). We find the integer value already taken and need a slightly
592 different one (let's call it the middle-case). The difference between
593 the two situations is that the next user will want to find something
594 higher than my number in the end-case and something lower than my number
597 So I suggest we give the function both a value and an upper bound and it
598 calculates us a primitive middle. The upper bound in the middle-case is
599 the next integer. The upper bound on the end-case is the nextup floating
600 point number. But the latter poses another problem: if we have occupied
601 the middle m between x and nextup(x), then the nextup(m) will probably
602 not be the same as nextup(x) because some rounding will take place
603 before the nextup is calculated and when the rounding reaches the
604 nextup(x), we will end up at nextup(nextup(x)).
606 So we really need to consider the nextup and the nextdown from there and
607 then the middle and that's the number we may approach asymptotically.
610 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
612 * consider deprecating the use of RECENT.recent as a symlink. It turns
613 out to need extra hoops with the rsync options and just isn't worth it.
614 Or maybe these extra hoops are needed anyway for the rest of the tree?
615 Nope, can't be the case because not all filesystems support symlinks.
617 But before doing the large step, I'll deprecate the call of
618 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
619 only introduced to resolve RECENT.recent and complicates the routine far
620 beyond what it deserves.
622 DONE. Won't deprecate RECENT.recent, just moved its handling to the
625 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
627 * I think it's a bug that the rsync_option links must be set to true in
628 order to support RECENT.recent and that nobody cares to set it
629 automatically. Similar for ignore_link_stat_errors.
631 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
633 * Todo: find all todos together and make a plan what is missing for a
636 - verifytree or something like that. fsck maybe.
638 - rersyncrecent, the script itself? What it do?
640 - a way to only mirror the recentfiles without mirroring the whole
641 remote system such that people can decide to mirror only partially see
642 also 2008-08-30. .shadow-xxx directory? this also needed for a
643 filesystem that is still incomplete and might need the mirrorfiles for
646 - long living objects that mirror again and again. Inject something
647 into ta, see how it goes over to tb.
649 - how do we continue filling up the DONE system when we use an object
650 for the second time? "fully covered" and "uptodate" or new terminology.
652 - overview called on the wrong file should be understandable
654 - the meta data field that must change when we fake something up so that
655 the downstream people know they have to re-fetch everything.
657 - how tolerant are we against missing files upstream? how do we keep
658 track? there are legitimate cases where we did read upstream index right
659 before a file got deleted there and then find that file as new and want
660 it. There are other cases that are not self healing and must be tracked
663 - how, exactly, do we have to deal with deletes? With rsync errors?
665 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
666 authors) failed: No such file or directory (2)
668 The file above is a delete in 1h and a new in file 1M and the
669 delete in the locally running rmirror did not get propagated to the 1M
670 object. Bug. And the consequence is a standstill.
672 It seems that a slave that works with a file below the principal needs
673 to merge things all the way up to get rid of later deletes. Or keep
674 track of all deletes and skip them later. So we need a trackdeletes.pm
675 similar to the done.pm?
677 see also 2008-08-20 about spurious deletes that really have no add
678 counterpart and yet they are not wrong.
680 - consider the effect when resyncing the recentfile takes longer than
681 the time per loop. Then we never rsync any file. We need to diagnose
682 that and force an increase of that loop time. But when we later are fast
683 enough again because the net has recovered, then we need to switch back
684 to original parameters. ERm, no, it's enough to keep syncing at least
685 one file before refetching an index file.
687 - remember to verify that no temp files are left lying around and the
690 - status file for not long running jobs that want to track upstream with
693 - revisit all XXX _float areas and study Sub::Exporter DONE
695 - persistent DB even though we just said we do not need it. Just for
696 extended capabilities and time savings when, for example, upstream
697 announces a reset and we get new recentfiles and could then limit
698 ourselves to a subset of files (those that have a changed epoch) in a
699 first pass and would only then do the loop to verify the rest. Or
702 * Todo: aggregate files should know their feed and finding the principal
703 should be done stepwise. (?)
705 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
706 left locks around. DONE
708 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
710 * maybe extend the _overview so that it always says if and where the
711 last file is in the next file and where the next event in the next rf
712 would lie. No, don't like this anymore. REJECT
714 * take the two new redundant tests out again, only the third must
717 * Todo: add a sanity check if the merged structure is really pointing to
718 a different rf and that this different rf is larger. DONE
720 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
722 * now test, if they are overlapping. And test if there is a file in the
723 next rf that would fit into this rf's interval.
725 1h 1222324012.8474 1222322541.7963 0.4086
726 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
727 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
728 1W 1222313218.3626 1221708477.5829 167.9835
730 I suspect that somebody writes a merged timestamp without having merged
731 and then somebody else relies on it.
733 If aggregate is running, the intervals must not be extravagated, if it
734 is not running, there must not be bounds, the total number of events in
735 the system must be counted and must be controlled throughout the tests.
736 That the test required the additional update was probably nonsense,
737 because aggregate can cut pieces too. FIXED & DONE
739 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
741 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
744 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
746 * Most apparent bug at the moment is that the recentfiles are fetched
747 too often. Only the principal should be fetched and if it has not
748 changed, the others should not be refetched. ATM I must admit that I'm
749 happy that we refetch more often than needed because I can more easily
750 fix bugs while the thing is running.
752 * Let's say, 1220474966.19501 is a timestamp of a file that is already
753 done but the done system does not know about it. The reason for the
754 failure is not known and we never reach the status uptodate because of
755 this. We must get over it.
757 Later it turns out that the origin server had a bug somewhere.
758 1220474966.19042 came after 1220474966.19501. Or better: it was in the
759 array of the recentfile one position above. The bug was my own.
761 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
763 * There is the race condition where the server does a delete and the
764 slave does not yet know and then tries to download it because he sees
765 the new. So for this time window we must be more tolerant against
766 failure. If we cannot download a file, we should just skip it and should
767 not retry immediately. The whole system should discover the lost thing
768 later. Keeping track with the DONE system should really be a no brainer.
770 But there is something more: the whole filesystem is a database and the
771 recentfiles are one possible representation of it. It's a pretty useful
772 representation I think that's why I have implemented something around
773 it. But for strictly local operation it has little value. For local
774 operation we would much rather have a database. So we would enter every
775 recentfile reading and every rsync operation and for every file the last
776 state change and what it leads to. Then we would always ignore older
777 records without the efforts involved with recentfiles.
779 The database would have: path,recentepoch,rsyncedon,deletedon
781 Oh well, not yet clear where this leads to.
783 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
785 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
786 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
788 * statusfile, maybe only on demand, alone to have a sharp debugging
789 tool. It is locked and all recentfiles dump themselves into it and we
790 can build a viewer that lets us know where we stand and what's inside.
792 * remember: only the principal recentfile needs expiration, all others
793 shall be expired by principal if it discovers that something has move
796 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
798 * Always check if we stringify to a higher value than in the entry
801 * And in covered make an additional check if we would be able to see a
802 numerical difference between the two numbers and if we can't then switch
803 to a different, more expensive algorithm. Do not want to be caught by
804 floating surprises. DONE
806 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
808 * caching has several aspects here: we can cache the interval of the
809 recentfile which only will change when the mtime of the file changes. We
810 must re-mirror the recentfile when its ttl has expired. Does have_read
811 tell you anything? It counts nothing at all. Only the mtime is
812 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
813 thing the have_mirrored because it is unrelated to the mtime.
815 * Robustness of floating point calculations! I always thought that the
816 string calculated by the origin server for the floating representation
817 of the epoch time is just a string. When we convert it to a number and
818 later back to a string, the other computer might come to a different
819 conclusion. This must not happen, we want to preserve it under any
820 circumstances. I will have to write tests with overlong sequences that
821 get lost in arithmetic and must see if all still works well. DONE
823 But one fragile point remains: if one host considers a>b and the other
824 one considers them == but no eq. To prevent this, we must probably do
825 some extra homework. DONE
827 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
829 * the concept of tracking DONE needs an object per recentfile that has
830 something like these methods:
832 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
834 covered() register() covered()
836 The unclear thing is how we translate points in time into intervals. We
837 could pass a reference to the current recent_events array when running
838 we_have(xxx) and let the DONE object iterate over it such that it only
839 has to store a list of intervals that can melt into each other. Ah, even
840 passing the list together with a list of indexes seems feasiable.
842 Or maybe ask for the inverted list?
844 Whenever the complete array is covered by the interval we say we are
845 fully covered and if the recentfile is not expired, we are uptodate.
847 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
849 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
851 * need a way to "return" the next entry after the end of a list. When
852 the caller says "before" or "after" we would like to know if he could
853 cover that interval/threshold or not because this influences the effect
854 of a newer timestamp of that recentfile. DONE with $opt{info}.
856 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
858 * one of the next things to tackle: the equivalent of csync2 -TIXU.
860 loop implies tixu (?). Nope, something like --statefile decides. Per
863 T test, I init, X including removals, U nodirtymark
865 So we have no concept of dirtymarks, we only trust that since we are
866 running we have observed everything steadily. But people will not let
867 this program run forever so we must consider both startup penalty and
868 book keeping for later runs. We keep this for later. For now we write a
869 long running mirror that merges several intervals.
871 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
873 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
876 * rersyncrecent, the script: default to one week. The name of the switch
877 is --after. Other switches? --loop!
879 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
881 * need a switch --skip-deletes (?)
883 * need a switch --enduser that tells us that the whole tempfile
884 discipline is not needed when there is no downstream user. (?)
886 Without this switch we cannot have a reasonable recent.pl that just
887 displays the recent additions. Either we accept to download everything.
888 Or we download temporary files without the typical rsync protocol
891 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
892 File::Temp::tempdir, this might be a win.
894 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
896 * apropos missing: we have no push, we never know the downstream
897 servers. People who know their downstream hosts and want to ascertain
898 something will want additional methods we have never thought about, like
899 update or delete a certain file.
901 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
903 * tempted to refactor rmirror into resolve_symlink, localize, etc.
904 Curious if rsync_options=links equal 0 vs. 1 will make the expected
907 * rsync options: it's a bit of a pain that we usually need several rsync
908 options, like compress, links, times, checksum and that there is no
909 reasonable default except the original rsync default. I think wee can
910 safely assume that the rsync options are shared between all recentfile
911 instances within one recent tree.
913 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
915 * deletes: if a delete follows an add quickly enough it may happen that
916 a downstream mirror did not see the add at all! It seems this needs to
917 be mentioned somewhere. The point here is that even if the downstream is
918 never missing the principal timeframe it may encounter a "delete" that
919 has no complimentary "add" anywhere.
921 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
923 * I suspect the treat of metadata is incorrect during read or something.
924 The bug that I am watching is that between 06:08 and 06:09 the 6h file
925 contained more than 6 hours worth of data. At 06:08 we merged into the
926 1d file. We need to take snapshots of the 6h file over the course of an
927 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
930 Much worse: watching the 1h file: right at the moment (at 06:35) it
931 covers 1218867584-1219120397 which is 70 hours.
933 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
934 08:19:44 2008, that is when I checked out last time, so it seems to be
935 aggregating and never truncating?
937 No, correct is: it is never truncating; but wrong is: it is aggregating.
938 It does receive a lot of events from time to time from a larger file.
939 Somehow a large file gets merged into the small one and because the
940 "meta/merged" attribute is missing, nobody is paying attention. I
941 believe that I can fix this by making sure that metadata are honoured
942 during read. DONE and test adjusted.
944 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
946 * grand renaming plan
948 remotebase => remoteroot to fit well with localroot DONE
949 local_path() => localroot seems to me should already work DONE
950 recentfile_basename => rfilename no need to stress it has no slash DONE
952 filenameroot??? Doesn't seem too bad to me today. Maybe something like
953 kern? It would anyway need a deprecation cycle because it is an
954 important constructor.
956 * I like the portability that Data::Serializer brings us but the price
957 is that some day we might find out that it is slowing us a bit. We'll
960 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
962 * should we not enter the interval of the principal (or the interval of
963 the merging file?) in every aggregated/merged file?
965 * we should aim at a first release and give up on thinking about
966 sanitizing stuff and zloop. Let's just admit that a full traditional
967 rsync is the only available sanitizer ATM. Otherwise it's complicated
968 stuff: sanitizing on the origin server, sanitizing on the slaves,
969 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
970 get the basics out before this becomes a major cause for mess.
972 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
974 * On OSes not supporting symlinks we expect that RECENT.recent contains
975 the contents of the principal recentfile. Actually this is identical on
976 systems supporting symlinks. Simple, what follows from that is that we
977 need to keep the serializer in the metadata because we cannot read it
978 from the filename, doesn't it? Of course not. It's a chicken and egg
979 problem. This leaves us with the problem to actually parse the
980 serialized data to find out in which format it is. So who can do the 4
981 or 5 magics we wanted to support? File::LibMagic?
983 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
985 * remotebase and recentfile_basename are ugly names. Now that we need a
986 word for the shortest/principal/driving recentfile too we should do
989 localroot is good. rfile is good. local_path() is bad, local_path($path)
990 is medium, filenameroot() is bad, remotebase is bad, recentfile is
993 Up to now remotebase was the string that described the remote root
994 directory in rsync notation, like pause.perl.org::authors. And
995 recentfile_basename was "RECENT-1h.yaml".
997 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
999 * The test that was added in today's checkin is a good start for a test
1000 of rmirror. We should have more methods in Recent.pm: verify,
1001 addmissingfiles. We should verify the current tree, then rmirror it and
1002 then verifytree the copy. We could then add some arbitrary file and let
1003 it be discovered by addmissingfiles, then rmirror again and then
1004 verifytree the copy again.
1006 Then we could start stealing from csync2 sqlite database [no port to
1007 OSX!] and fill a local DB. And methods to compare the database with the
1008 recentfiles. Our strength is that in principle we could maintain state
1009 with a single float. We have synced up to 1234567890.123456. If the Z
1010 file does not add new files all we have to do is mirror the new ones and
1013 This makes it clear that we should extend current protocol and declare
1014 that we cheat when we add files too late, just to help the other end
1015 keeping track. Ah yes, that's what was meant when zloop was mentioned
1018 Maybe need to revisit File::Mirror to help me with this task.
1020 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1022 * There must be an allow-me-to-truncate flag in every recentfile.
1023 Without it one could construct a sequence of updates winning the locking
1024 battle against the aggregator. Only if an aggregator has managed to
1025 merge data over to the next level, truncating can be allowed. DONE with
1028 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1030 * We should probably guarantee that no duplicates enter the aggregator
1033 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1035 * To get merge operation faster would need a good benchmark test. What
1036 02 spits out isn't reliable enough and is dominated by many other
1039 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1040 Author: Andreas J. Koenig <k@k75.(none)>
1041 Date: Sat Aug 2 07:58:04 2008 +0200
1045 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1046 Author: Andreas J. Koenig <k@k75.(none)>
1047 Date: Sat Aug 2 11:40:29 2008 +0200
1049 there was a lot of trying but the effect is hardly measurable with
1052 * overhead of connecting seems high. When setting
1053 max_files_per_connection to 1 we see that.
1055 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1057 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1059 25138.2215600014/3600 = 6.98283932222261
1061 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1062 be a bug in the aggregator. FIXED
1064 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1066 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1067 we have it, can we mirror it, mirror it!
1069 I fear this needs a new class which might be called
1070 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1071 and treat them as an entity. I realize that a single recentfile may be
1072 sufficient for certain tasks and that it is handy for the low level
1073 programmer but it is not nice to use. If there is a delete in the 1h
1074 file then the 6h file still contains it. Seekers of the best information
1075 need to combine at least some of the recentfiles most of the time.
1077 There is the place for the Z loop!
1079 But the combination is something to collect in a database, isn't it. Did
1080 csync2 just harrumph?
1082 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1084 * it just occurred to me that hosts in the same mirroring pool could
1085 help out each other even without rewriting the recentfile. Just fetch
1086 the stuff to mirror from several places, bingo. But that's something
1087 that should rather live in a separate package or in rsync directly.
1089 * cronjobs are unsuited because with ntp they would all come at the full
1090 minute and disturb each other. Besides that I'd hate to have a backbone
1091 with more than a few seconds latency.
1093 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1095 * a second rsync server with access control for PAUSE. Port? 873 is the
1096 standard port, let's take 8873.
1098 * if there were a filesystem based on this, it would have a slow access
1099 to inexistent files. It would probably provide wrong readdir (only based
1100 on current content) or also a slow one (based on a recentfile written
1101 after the call). But it would provide fast access to existing files. Or
1102 one would deliberately allow slightly blurred answers based on some
1103 sqlite reflection of the recentfiles.
1105 * todo: write a variant of mirror() that combines two or more
1106 recentfiles and treats them like one
1108 * todo: signal handler to remove the tempfile
1110 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1112 * now that we have the symlink I forgot how it should be used in
1115 * the z loop: add missing files to Z file. Just append them (instead of
1116 prepending). So one guy prepends something from the Y file from time to
1117 time and another guy appends something rather frequently. Collecting
1118 pond. When Y merges into Z, things get epoch and the collecting pond
1119 gets smaller. What exactly are "missing files"?
1121 take note of current epoch of the alpha file, let's call it the
1124 find all files on disk
1126 remove all files registered in the recentworld up to recent-ts
1128 remove all files that have been deleted after recent-ts according to
1131 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1133 * rersyncrecent might be a cronjob with a (locked) state file which
1134 contains things like after and maybe last z sync or such?
1136 rrr-mirror might be an alternative name but how would we justify the
1137 three Rs when there is no Re-Rsync-Recent?
1139 With the --loop parameter it is an endless loop, without it is no loop.
1140 At least this is simple.
1142 * todo: new accssor z-interval specifies how often the Z file is updated
1143 against the filesystem. We probably want no epoch stamp on these
1144 entries. And we want to be able to filter the entries (e.g. no
1145 by-modules and by-category tree)
1147 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1149 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1150 into Z. We do not need the other files filled up with filesystem
1153 * need interface to query for a file in order to NOT call update on
1154 PAUSE a second time within a short time.
1156 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1158 * recommended update interval? Makes no sense, is different for
1165 change-log-default-name: "Todo"