1 2009-03-22 Andreas J. Koenig <andk@cpan.org>
3 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
4 about it all the time and nobody reacts accordingly; (2) lots of temp
5 files get created and not removed; culprit the new call to
6 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
7 callere has to remove the tempfile after use.
9 Need the drawing board.
11 2009-03-21 Andreas J. Koenig <andk@cpan.org>
13 * Bug?: should it be harder than it is atm to set the timestamp to the
16 * bug with native integers:
20 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
24 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
27 Native integer broke when native math was turned off. FIXED
29 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
30 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
33 Yes, records out of order:
36 447004 epoch: 995885533
37 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
40 447008 epoch: 995890358
41 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
44 447012 epoch: 995892221
45 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
48 FIXED with sanity check and later with the integer fix.
50 * Bug: want the index files in a .recent directory
52 * Bug: lots of dot files are not deleted in time
54 * possible test case: can a delete change the timestamp? This would
55 probably break the order of events.
57 2009-03-20 Andreas J. Koenig <andk@cpan.org>
59 * 1233701831.34486 what's so special about this number/string? It
61 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
62 atm lives in Y,Q, and Z.
64 It is the first entry after id/--skip-locking which has timestamp
65 1234164228.11325 which represents a file that doesn't exist anymore.
69 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
70 _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
71 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
72 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
73 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
74 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
75 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
76 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
77 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
78 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
79 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
80 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
81 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
83 FIXED, it was the "--skip-locking" file where manual intervention was participating
85 * bug on the mirroring slave: when the dirtymark gets increased we
86 probably do not reset the done intervals. The mirrorer stays within
87 tight bounds where it tries to sync with upstream and never seems to
88 finish. In the debugging state file I see lots of identical intervals
89 that do not get collapsed. When I restart the mirrorer it dies with:
91 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
92 _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
93 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
94 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
95 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
96 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
97 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
98 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
99 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
100 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
102 and debugging stands at
108 epoch: 1237400802.5789
109 path: id/J/JO/JOHND/CHECKSUMS
112 epoch: 1237400807.97514
113 path: id/M/MI/MIYAGAWA/CHECKSUMS
116 epoch: 1237400817.94363
117 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
125 and it is reproducable.
127 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
128 pause has a fresh one.
130 FIXED, it was the third anded term in each of the ifs in the IV block in
131 _register_one: with that we make sure that we do not stamp on valuable
134 2009-03-17 Andreas J. Koenig <andk@cpan.org>
136 * done: verified the existence of the floating point bug in bleadperl
137 and verified that switching from YAML::Syck to YAML::XS does not resolve
140 BTW, the switch was doable with
142 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
144 and should be considered as a separate TODO
146 * todo: integrate a dirty update with two aggregate calls before
147 unlocking for frictionless dirtying
149 * todo: start the second rsync daemon on pause
151 * todo: move index files to .recent: this cannot simply be done by
152 setting filenameroot to .recent/RECENT. Other parts of the modules rely
153 on the fact that dirname(recentfile) is the root of the mirrored tree.
155 2009-03-16 Andreas J. Koenig <andk@cpan.org>
157 * What was the resolution of the mirror.pl delete hook bug? Do we call
158 the delete hook when pause removes a file from MUIR?
160 * Today on pause: Updating 2a13fba..29f284d and installing it for
161 /usr/local/perl-5.10.0{,-RC2}
163 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
164 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
166 2009-03-15 Andreas J. Koenig <andk@cpan.org>
168 * currently recent_events has the side effect of setting dirtymark
169 because it forces a read on the file. That should be transparent, so
170 that the dirtymark call always forces a cache-able(?) read.
172 * The bug below is -- after a lot of trying -- not reproducible on a
173 small script, only in the large test script. The closest to the output
179 my $x = "01237123229.8814";
184 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
186 $l = "1237123231.22458";
187 $r = "1237123231.22458";
189 Devel::Peek::Dump $l;
190 Devel::Peek::Dump $r;
191 Devel::Peek::Dump $x = $l <=> $r;
195 The checked in state at c404a85 fails the test with my
196 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
197 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
198 in the test script always consistent.
200 * Todo: write a test that inserts a second dirty file with an already
201 existing timestamp. DONE
203 * Bug in perl 5.10 on my 64bit box:
205 DB<98> Devel::Peek::Dump $l
206 SV = PVMG(0x19e0450) at 0x142a550
208 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
210 NV = 1237123231.22458
211 PV = 0x194ce70 "1237123231.22458"\0
215 DB<99> Devel::Peek::Dump $r
216 SV = PVMG(0x19e0240) at 0x142a3e8
218 FLAGS = (PADMY,POK,pPOK)
221 PV = 0x19ff900 "1237123231.22458"\0
225 DB<100> Devel::Peek::Dump $l <=> $r
226 SV = IV(0x19ea6e8) at 0x19ea6f0
228 FLAGS = (PADTMP,IOK,pIOK)
231 DB<101> Devel::Peek::Dump $l
232 SV = PVMG(0x19e0450) at 0x142a550
234 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
236 NV = 1237123231.22458
237 PV = 0x194ce70 "1237123231.22458"\0
241 DB<102> Devel::Peek::Dump $r
242 SV = PVMG(0x19e0240) at 0x142a3e8
244 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
246 NV = 1237123231.22458
247 PV = 0x19ff900 "1237123231.22458"\0
251 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
253 * Todo: reset "done" or "covered" and "minmax" after a dirty operation?
255 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
257 * $obj->merge ($other) needs to learn about equal epoch which may happen
258 since dirty_epoch intruded.
260 * Wontfix anytime soon: I think we currently do not support mkdir. Only
263 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
265 * Todo: continue working on update(...,$dirty_epoch). It must be
266 followed by a fast_aggregate!
268 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
270 * maybe we need a closest_entry or fitting_interval or something like
271 that. We want to merge an event into the middle of some recentfile.
272 First we do not know which file, then we do not know where to lock,
273 where to enter the new item, when and where to correct the dirtymark.
275 So my thought is we should first find which file.
277 Another part of my brain answers: what would happen if we would enter
278 the new file into the smallest file just like an ordinary new event,
279 just as an old event?
281 (1) we would write a duplicate timestamp? No, this would be easy to
284 (2) we would make the file large quickly? Yes, but so what? We are
285 changing the dirtymark, so are willing to disturb the downstream hosts.
287 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
289 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
293 https://rt.cpan.org/Ticket/Display.html?id=41199
295 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
297 Strangest output being HAKANARDO who managed to upload
299 Here is a better oneliner that includes also the first line of each
302 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
304 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
306 * memory leak in the syncher? It currently weighs 100M.
310 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
313 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
315 * lookup by epoch and by path and use this ability on the pause to never
316 again register a file twice that doesn't need it. Let's call it
319 * after the dirtymark is done: fill up recentfiles with fake (historic)
320 entries; fill up with individual corrections; algorithm maybe to be done
321 with bigfloat so that we can always place something in the middle
322 between two entries. Before we must switch to bigfloat we could try to
323 use Data::Float::nextup to get the.
325 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
328 * dirtymark now lives in Recentfile, needs to be used in rmirror.
330 * find out why the downloader died after a couple of hours without a net
331 connection. Write a test that survives the not-existence of the other
334 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
336 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
337 it makes less sense with tight coupling for secondary files. But asking
338 the server what the current epoch is might be cheaper on HTTP than on
339 rsync. (Needs to be evaluated)
341 * remove the 0.00 from the verbose overview in the Merged column in the
344 * write tests that expose the problems of the last few days: cascading
345 client/server roles, tight coupling for secondary RFs, deletes after
348 * Some day we might want to have policy options for the slave:
349 tight/loose/no coupling with upstream for secondary RFs. tight is what
350 we have now. loose would wait until a gap occurs that can be closed.
352 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
354 * revisit all $rfs->[$i+1] places if they now make sense still
356 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
358 * another bug is the fact that the mirror command deletes files before
359 it unhides the index file, thus confusing downstream slaves. We must not
360 delete before unhiding and must delete after unhiding. FIXED.
362 * new complication about the slave that is playing a server role.
363 Currently we mirror from newest to oldest with a hidden temporary file
364 as index. And when one file is finished, we unhide the index file.
365 Imagine the cascading server/slave is dead for a day. It then starts
366 mirroring again with the freshest thing and unhides the freshest index
367 file when it has worked through it. In that moment it exposes a time
368 hole. Because it now works on the second recentfile which is still
371 We currently do nothing special to converge after such a drop out. At
372 least not intentionally and robustly and thought through.
374 The algorithm we use to seed the next file needs quite a lot of more
375 robustness than it currently has. Something to do with looking at the
376 merged element of the next rf and when it has dropped off, we seed
377 immediately. And if it ramains dropped off, we seed again, of course.
379 Nope, looking from smaller to larger RFS we look at the merged element
380 of this RF and at the minmax/max element of the next RF. If that
381 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
384 And we need a public accessor seed and unseed or seeded. But not the mix
385 of public and private stuff that then is used behind the back.
387 And then the secondary* stuff must go.
389 And we must understand what the impact is on the DONE system. Can it go
390 unnoticed that there was a hole? And could the DONE system have decided
391 the hole is covered? This should be testable with three directories where
392 the middle stops working for a while. Done->merge is suspicious, we must
393 stop it from merging non-conflatable neighbors due to broken continuity.
397 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
399 * Slaven suggests to have the current epoch or the whole current
400 recentfile available from the HTTP server and take it away with
401 keepalive. This direction goes the granularity down to subseconds.
403 We might want to rewrite everything to factor out transport and allow
404 the whole thing to run via HTTP.
406 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
408 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
411 * smoker on k81 fetching from k75 to verify cascading works. See
412 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
414 * maybe the loop should wait for CHECKSUMS file after every upload. And
415 CPAN.pm needs to deal with timestamps in the future.
417 * do not forget the dirtymark!
419 Text: have a new flag on recentfiles with the meaning: if this
420 changes, you're required to run a full rsync over all the files. The
421 reason why we set it would probably be: some foul happened. we injected
422 files in arbitrary places or didn't inject them although they changed.
423 The content of the flag? Timestamp? The relation between the
424 recentfiles would have to be inheritance from the principal, because any
425 out of band changes would soon later propagate to the next recentfile.
427 By upping the flag often one can easily ruin the slaves.
429 last out of band change? dirtymark?
431 Anyway, this implies that we read a potentially existing recentfile
434 And it implies that we have an eventloop that keeps us busy in 2-3
435 cycles, one for current stuff (tight loop) and one for the recentfiles
436 (cascade when principal has changed), one for the old stuff after a
439 And it implies that the out-of-band change in any of the recentfiles
440 must have a lock on the principal file and there is the place to set the
443 * start a FAQ, especially quick start guide questions. Also to aid those
444 problematic areas where we have no good solution, like the "links"
447 * wish feedback when we are slow.
451 * Remove a few DEBUG statements.
453 * The multiple-rrr way of doing things needs a new option to rmirror,
454 like piecemeal or so. Not urgent because after the first pass through,
455 things run smoothely. It's only ugly during the first pass.
457 * I have the suspicion that the code is broken that decides if the
458 neighboring RF needs to be seeded. I fear when too much time has gone
459 between two calls (in our case more than one hour), it would not seed
460 the neighbor. Of course this will never be noticed, so we need a good
463 * local/localroot confusion: I currently pass both options but one must
466 * accounts for early birds on PAUSE rsync daemon.
468 * hardcoded 20 seconds
470 * who mirrors the index? DOING now.
472 * which CPAN mirrors offer rsync?
474 * visit all XXX, visit all _float places
476 * rename the pathdb stuff, it's too confusing. No idea how.
478 * rrr-inotify, backpan, rrr-register
480 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
482 * current bugs: the pathdb seems to get no reset, the seeding of the
483 secondaryttl stuff seems not to have an effect. Have helped myself with
484 a rand(10), need to fix this back. So not checked in. Does the rand
487 The rand thing helps. The secondaryttl stuff was in the wrong line,
490 The pathdb stuff was because I called either _pathdb or __pathdb on the
491 wrong object. FIXED now.
493 * It's not so beautiful if we never fetch the recentfiles that are not
494 the principal, even if this is correct behaviour. We really do not need
495 them after we have fetched the whole content.
497 OK, we want a switch for that: secondaryttl DONE
499 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
501 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
502 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
503 it lets 40 follow. FIXED
505 * See that the long running process really only updates the principal
506 file unless it has missed a timespan during which something happened. If
507 nothing happened, it must notice even when it misses the timespan. DONE
509 * we must throw away the pathdb when we have reached the end of Z. From
510 that moment we can have a very small pathdb because the only reason for
511 a pathdb is that we know to ignore old records in old files. We won't
512 need this pathdb again before the next full pass over the data is
513 necessary and then we will rebuild it as we go along. DONE
515 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
517 * I think, Done::register_one is doing wrong in that it does not
518 conflate neighboring pieces. The covered() method cannot do this because
519 it has no recent_events array at hand. But register_one has it and could
520 do it and for some reason misses to do it (sometimes).
522 This means that the three tests I just wrote can probably not survive
523 because they test with an already broken Done structure.
525 The art now is to detect how it happens, then to reproduce, then write a
528 So from the logfile this is what happens: we have a good interval with
529 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
530 top of it. Locally we now mirror F2 and open a new done interval for it.
531 Then we mirror F1 but this time with the timestamp T1b. And when we then
532 try to close the gap, we do not find T1 but instead something older. We
533 should gladly accept this older piece and this would fix this bug.
537 * bug to fix: when the 1h file changes while rmirror is running, we do
538 correctly sync the new files but never switch to the 6h file but rather
539 stay in a rather quick loop that fetches the 1h file again and again.
541 Is it possible that we initialize a new object? Or does
542 get_remote_recentfile_as_tempfile overwrite something in myself?
544 Want a new option: _runstatusfile => $file which frequently dumps the
545 state of all recentfiles to a file.
549 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
551 * Todo: now teach update to verify the timestamp is about to write
552 against the previous and use _increase_a_bit if it doesn't comply with
553 strict monotony. DONE
555 * The problem of rounding. So far perl's default precision was
556 sufficient. One day it won't be. FakeFloat has an easy job when it is
557 only reading and other machines have written correctly. But when we want
558 to write a floating point number that is a bit larger than the other
559 one, then we need our own idea of precision.
561 Slaven said: just append a "1". This might be going towards the end of
562 usability too quickly. I'd like something that actually uses the decimal
563 system. Well, appending a 1 also does this but...
565 E.g. we have 1.0. nextup on this architecture is starting with
566 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
567 taken the 1.0000000000000003 and the next user comes and the time tells
568 him 1.0 again. He has to beat my number without stepping over the
569 nextup. This is much less space than I had when I chose 1,2,3.
571 What is also irritating is that nextup is architecture dependent. The
572 128 bit guy must choose very long numbers to fit in between whereas the
573 other one with 16 bit uses larger steps. But then the algorithm is the
574 same for both, so that would be a nice thing.
576 I see two situation where we need this. One is when Time::HiRes returns
577 us a value that is <= the last entry in our recentfile. In this case
578 (let's call it the end-case) we must fill the region between that number
579 and the next higher native floating point number. The other is when we
580 inject an old file into an old recentfile (we would then also set a new
581 dirtymark). We find the integer value already taken and need a slightly
582 different one (let's call it the middle-case). The difference between
583 the two situations is that the next user will want to find something
584 higher than my number in the end-case and something lower than my number
587 So I suggest we give the function both a value and an upper bound and it
588 calculates us a primitive middle. The upper bound in the middle-case is
589 the next integer. The upper bound on the end-case is the nextup floating
590 point number. But the latter poses another problem: if we have occupied
591 the middle m between x and nextup(x), then the nextup(m) will probably
592 not be the same as nextup(x) because some rounding will take place
593 before the nextup is calculated and when the rounding reaches the
594 nextup(x), we will end up at nextup(nextup(x)).
596 So we really need to consider the nextup and the nextdown from there and
597 then the middle and that's the number we may approach asymptotically.
600 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
602 * consider deprecating the use of RECENT.recent as a symlink. It turns
603 out to need extra hoops with the rsync options and just isn't worth it.
604 Or maybe these extra hoops are needed anyway for the rest of the tree?
605 Nope, can't be the case because not all filesystems support symlinks.
607 But before doing the large step, I'll deprecate the call of
608 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
609 only introduced to resolve RECENT.recent and complicates the routine far
610 beyond what it deserves.
612 DONE. Won't deprecate RECENT.recent, just moved its handling to the
615 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
617 * I think it's a bug that the rsync_option links must be set to true in
618 order to support RECENT.recent and that nobody cares to set it
619 automatically. Similar for ignore_link_stat_errors.
621 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
623 * Todo: find all todos together and make a plan what is missing for a
626 - verifytree or something like that. fsck maybe.
628 - rersyncrecent, the script itself? What it do?
630 - a way to only mirror the recentfiles without mirroring the whole
631 remote system such that people can decide to mirror only partially see
632 also 2008-08-30. .shadow-xxx directory? this also needed for a
633 filesystem that is still incomplete and might need the mirrorfiles for
636 - long living objects that mirror again and again. Inject something
637 into ta, see how it goes over to tb.
639 - how do we continue filling up the DONE system when we use an object
640 for the second time? "fully covered" and "uptodate" or new terminology.
642 - overview called on the wrong file should be understandable
644 - the meta data field that must change when we fake something up so that
645 the downstream people know they have to re-fetch everything.
647 - how tolerant are we against missing files upstream? how do we keep
648 track? there are legitimate cases where we did read upstream index right
649 before a file got deleted there and then find that file as new and want
650 it. There are other cases that are not self healing and must be tracked
653 - how, exactly, do we have to deal with deletes? With rsync errors?
655 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
656 authors) failed: No such file or directory (2)
658 The file above is a delete in 1h and a new in file 1M and the
659 delete in the locally running rmirror did not get propagated to the 1M
660 object. Bug. And the consequence is a standstill.
662 It seems that a slave that works with a file below the principal needs
663 to merge things all the way up to get rid of later deletes. Or keep
664 track of all deletes and skip them later. So we need a trackdeletes.pm
665 similar to the done.pm?
667 see also 2008-08-20 about spurious deletes that really have no add
668 counterpart and yet they are not wrong.
670 - consider the effect when resyncing the recentfile takes longer than
671 the time per loop. Then we never rsync any file. We need to diagnose
672 that and force an increase of that loop time. But when we later are fast
673 enough again because the net has recovered, then we need to switch back
674 to original parameters. ERm, no, it's enough to keep syncing at least
675 one file before refetching an index file.
677 - remember to verify that no temp files are left lying around and the
680 - status file for not long running jobs that want to track upstream with
683 - revisit all XXX _float areas and study Sub::Exporter DONE
685 - persistent DB even though we just said we do not need it. Just for
686 extended capabilities and time savings when, for example, upstream
687 announces a reset and we get new recentfiles and could then limit
688 ourselves to a subset of files (those that have a changed epoch) in a
689 first pass and would only then do the loop to verify the rest. Or
692 * Todo: aggregate files should know their feed and finding the principal
693 should be done stepwise. (?)
695 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
696 left locks around. DONE
698 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
700 * maybe extend the _overview so that it always says if and where the
701 last file is in the next file and where the next event in the next rf
702 would lie. No, don't like this anymore. REJECT
704 * take the two new redundant tests out again, only the third must
707 * Todo: add a sanity check if the merged structure is really pointing to
708 a different rf and that this different rf is larger. DONE
710 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
712 * now test, if they are overlapping. And test if there is a file in the
713 next rf that would fit into this rf's interval.
715 1h 1222324012.8474 1222322541.7963 0.4086
716 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
717 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
718 1W 1222313218.3626 1221708477.5829 167.9835
720 I suspect that somebody writes a merged timestamp without having merged
721 and then somebody else relies on it.
723 If aggregate is running, the intervals must not be extravagated, if it
724 is not running, there must not be bounds, the total number of events in
725 the system must be counted and must be controlled throughout the tests.
726 That the test required the additional update was probably nonsense,
727 because aggregate can cut pieces too. FIXED & DONE
729 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
731 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
734 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
736 * Most apparent bug at the moment is that the recentfiles are fetched
737 too often. Only the principal should be fetched and if it has not
738 changed, the others should not be refetched. ATM I must admit that I'm
739 happy that we refetch more often than needed because I can more easily
740 fix bugs while the thing is running.
742 * Let's say, 1220474966.19501 is a timestamp of a file that is already
743 done but the done system does not know about it. The reason for the
744 failure is not known and we never reach the status uptodate because of
745 this. We must get over it.
747 Later it turns out that the origin server had a bug somewhere.
748 1220474966.19042 came after 1220474966.19501. Or better: it was in the
749 array of the recentfile one position above. The bug was my own.
751 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
753 * There is the race condition where the server does a delete and the
754 slave does not yet know and then tries to download it because he sees
755 the new. So for this time window we must be more tolerant against
756 failure. If we cannot download a file, we should just skip it and should
757 not retry immediately. The whole system should discover the lost thing
758 later. Keeping track with the DONE system should really be a no brainer.
760 But there is something more: the whole filesystem is a database and the
761 recentfiles are one possible representation of it. It's a pretty useful
762 representation I think that's why I have implemented something around
763 it. But for strictly local operation it has little value. For local
764 operation we would much rather have a database. So we would enter every
765 recentfile reading and every rsync operation and for every file the last
766 state change and what it leads to. Then we would always ignore older
767 records without the efforts involved with recentfiles.
769 The database would have: path,recentepoch,rsyncedon,deletedon
771 Oh well, not yet clear where this leads to.
773 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
775 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
776 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
778 * statusfile, maybe only on demand, alone to have a sharp debugging
779 tool. It is locked and all recentfiles dump themselves into it and we
780 can build a viewer that lets us know where we stand and what's inside.
782 * remember: only the principal recentfile needs expiration, all others
783 shall be expired by principal if it discovers that something has move
786 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
788 * Always check if we stringify to a higher value than in the entry
791 * And in covered make an additional check if we would be able to see a
792 numerical difference between the two numbers and if we can't then switch
793 to a different, more expensive algorithm. Do not want to be caught by
794 floating surprises. DONE
796 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
798 * caching has several aspects here: we can cache the interval of the
799 recentfile which only will change when the mtime of the file changes. We
800 must re-mirror the recentfile when its ttl has expired. Does have_read
801 tell you anything? It counts nothing at all. Only the mtime is
802 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
803 thing the have_mirrored because it is unrelated to the mtime.
805 * Robustness of floating point calculations! I always thought that the
806 string calculated by the origin server for the floating representation
807 of the epoch time is just a string. When we convert it to a number and
808 later back to a string, the other computer might come to a different
809 conclusion. This must not happen, we want to preserve it under any
810 circumstances. I will have to write tests with overlong sequences that
811 get lost in arithmetic and must see if all still works well. DONE
813 But one fragile point remains: if one host considers a>b and the other
814 one considers them == but no eq. To prevent this, we must probably do
815 some extra homework. DONE
817 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
819 * the concept of tracking DONE needs an object per recentfile that has
820 something like these methods:
822 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
824 covered() register() covered()
826 The unclear thing is how we translate points in time into intervals. We
827 could pass a reference to the current recent_events array when running
828 we_have(xxx) and let the DONE object iterate over it such that it only
829 has to store a list of intervals that can melt into each other. Ah, even
830 passing the list together with a list of indexes seems feasiable.
832 Or maybe ask for the inverted list?
834 Whenever the complete array is covered by the interval we say we are
835 fully covered and if the recentfile is not expired, we are uptodate.
837 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
839 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
841 * need a way to "return" the next entry after the end of a list. When
842 the caller says "before" or "after" we would like to know if he could
843 cover that interval/threshold or not because this influences the effect
844 of a newer timestamp of that recentfile. DONE with $opt{info}.
846 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
848 * one of the next things to tackle: the equivalent of csync2 -TIXU.
850 loop implies tixu (?). Nope, something like --statefile decides. Per
853 T test, I init, X including removals, U nodirtymark
855 So we have no concept of dirtymarks, we only trust that since we are
856 running we have observed everything steadily. But people will not let
857 this program run forever so we must consider both startup penalty and
858 book keeping for later runs. We keep this for later. For now we write a
859 long running mirror that merges several intervals.
861 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
863 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
866 * rersyncrecent, the script: default to one week. The name of the switch
867 is --after. Other switches? --loop!
869 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
871 * need a switch --skip-deletes (?)
873 * need a switch --enduser that tells us that the whole tempfile
874 discipline is not needed when there is no downstream user. (?)
876 Without this switch we cannot have a reasonable recent.pl that just
877 displays the recent additions. Either we accept to download everything.
878 Or we download temporary files without the typical rsync protocol
881 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
882 File::Temp::tempdir, this might be a win.
884 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
886 * apropos missing: we have no push, we never know the downstream
887 servers. People who know their downstream hosts and want to ascertain
888 something will want additional methods we have never thought about, like
889 update or delete a certain file.
891 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
893 * tempted to refactor rmirror into resolve_symlink, localize, etc.
894 Curious if rsync_options=links equal 0 vs. 1 will make the expected
897 * rsync options: it's a bit of a pain that we usually need several rsync
898 options, like compress, links, times, checksum and that there is no
899 reasonable default except the original rsync default. I think wee can
900 safely assume that the rsync options are shared between all recentfile
901 instances within one recent tree.
903 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
905 * deletes: if a delete follows an add quickly enough it may happen that
906 a downstream mirror did not see the add at all! It seems this needs to
907 be mentioned somewhere. The point here is that even if the downstream is
908 never missing the principal timeframe it may encounter a "delete" that
909 has no complimentary "add" anywhere.
911 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
913 * I suspect the treat of metadata is incorrect during read or something.
914 The bug that I am watching is that between 06:08 and 06:09 the 6h file
915 contained more than 6 hours worth of data. At 06:08 we merged into the
916 1d file. We need to take snapshots of the 6h file over the course of an
917 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
920 Much worse: watching the 1h file: right at the moment (at 06:35) it
921 covers 1218867584-1219120397 which is 70 hours.
923 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
924 08:19:44 2008, that is when I checked out last time, so it seems to be
925 aggregating and never truncating?
927 No, correct is: it is never truncating; but wrong is: it is aggregating.
928 It does receive a lot of events from time to time from a larger file.
929 Somehow a large file gets merged into the small one and because the
930 "meta/merged" attribute is missing, nobody is paying attention. I
931 believe that I can fix this by making sure that metadata are honoured
932 during read. DONE and test adjusted.
934 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
936 * grand renaming plan
938 remotebase => remoteroot to fit well with localroot DONE
939 local_path() => localroot seems to me should already work DONE
940 recentfile_basename => rfilename no need to stress it has no slash DONE
942 filenameroot??? Doesn't seem too bad to me today. Maybe something like
943 kern? It would anyway need a deprecation cycle because it is an
944 important constructor.
946 * I like the portability that Data::Serializer brings us but the price
947 is that some day we might find out that it is slowing us a bit. We'll
950 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
952 * should we not enter the interval of the principal (or the interval of
953 the merging file?) in every aggregated/merged file?
955 * we should aim at a first release and give up on thinking about
956 sanitizing stuff and zloop. Let's just admit that a full traditional
957 rsync is the only available sanitizer ATM. Otherwise it's complicated
958 stuff: sanitizing on the origin server, sanitizing on the slaves,
959 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
960 get the basics out before this becomes a major cause for mess.
962 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
964 * On OSes not supporting symlinks we expect that RECENT.recent contains
965 the contents of the principal recentfile. Actually this is identical on
966 systems supporting symlinks. Simple, what follows from that is that we
967 need to keep the serializer in the metadata because we cannot read it
968 from the filename, doesn't it? Of course not. It's a chicken and egg
969 problem. This leaves us with the problem to actually parse the
970 serialized data to find out in which format it is. So who can do the 4
971 or 5 magics we wanted to support? File::LibMagic?
973 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
975 * remotebase and recentfile_basename are ugly names. Now that we need a
976 word for the shortest/principal/driving recentfile too we should do
979 localroot is good. rfile is good. local_path() is bad, local_path($path)
980 is medium, filenameroot() is bad, remotebase is bad, recentfile is
983 Up to now remotebase was the string that described the remote root
984 directory in rsync notation, like pause.perl.org::authors. And
985 recentfile_basename was "RECENT-1h.yaml".
987 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
989 * The test that was added in today's checkin is a good start for a test
990 of rmirror. We should have more methods in Recent.pm: verify,
991 addmissingfiles. We should verify the current tree, then rmirror it and
992 then verifytree the copy. We could then add some arbitrary file and let
993 it be discovered by addmissingfiles, then rmirror again and then
994 verifytree the copy again.
996 Then we could start stealing from csync2 sqlite database [no port to
997 OSX!] and fill a local DB. And methods to compare the database with the
998 recentfiles. Our strength is that in principle we could maintain state
999 with a single float. We have synced up to 1234567890.123456. If the Z
1000 file does not add new files all we have to do is mirror the new ones and
1003 This makes it clear that we should extend current protocol and declare
1004 that we cheat when we add files too late, just to help the other end
1005 keeping track. Ah yes, that's what was meant when zloop was mentioned
1008 Maybe need to revisit File::Mirror to help me with this task.
1010 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1012 * There must be an allow-me-to-truncate flag in every recentfile.
1013 Without it one could construct a sequence of updates winning the locking
1014 battle against the aggregator. Only if an aggregator has managed to
1015 merge data over to the next level, truncating can be allowed. DONE with
1018 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1020 * We should probably guarantee that no duplicates enter the aggregator
1023 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1025 * To get merge operation faster would need a good benchmark test. What
1026 02 spits out isn't reliable enough and is dominated by many other
1029 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1030 Author: Andreas J. Koenig <k@k75.(none)>
1031 Date: Sat Aug 2 07:58:04 2008 +0200
1035 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1036 Author: Andreas J. Koenig <k@k75.(none)>
1037 Date: Sat Aug 2 11:40:29 2008 +0200
1039 there was a lot of trying but the effect is hardly measurable with
1042 * overhead of connecting seems high. When setting
1043 max_files_per_connection to 1 we see that.
1045 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1047 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1049 25138.2215600014/3600 = 6.98283932222261
1051 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1052 be a bug in the aggregator. FIXED
1054 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1056 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1057 we have it, can we mirror it, mirror it!
1059 I fear this needs a new class which might be called
1060 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1061 and treat them as an entity. I realize that a single recentfile may be
1062 sufficient for certain tasks and that it is handy for the low level
1063 programmer but it is not nice to use. If there is a delete in the 1h
1064 file then the 6h file still contains it. Seekers of the best information
1065 need to combine at least some of the recentfiles most of the time.
1067 There is the place for the Z loop!
1069 But the combination is something to collect in a database, isn't it. Did
1070 csync2 just harrumph?
1072 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1074 * it just occurred to me that hosts in the same mirroring pool could
1075 help out each other even without rewriting the recentfile. Just fetch
1076 the stuff to mirror from several places, bingo. But that's something
1077 that should rather live in a separate package or in rsync directly.
1079 * cronjobs are unsuited because with ntp they would all come at the full
1080 minute and disturb each other. Besides that I'd hate to have a backbone
1081 with more than a few seconds latency.
1083 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1085 * a second rsync server with access control for PAUSE. Port? 873 is the
1086 standard port, let's take 8873.
1088 * if there were a filesystem based on this, it would have a slow access
1089 to inexistent files. It would probably provide wrong readdir (only based
1090 on current content) or also a slow one (based on a recentfile written
1091 after the call). But it would provide fast access to existing files. Or
1092 one would deliberately allow slightly blurred answers based on some
1093 sqlite reflection of the recentfiles.
1095 * todo: write a variant of mirror() that combines two or more
1096 recentfiles and treats them like one
1098 * todo: signal handler to remove the tempfile
1100 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1102 * now that we have the symlink I forgot how it should be used in
1105 * the z loop: add missing files to Z file. Just append them (instead of
1106 prepending). So one guy prepends something from the Y file from time to
1107 time and another guy appends something rather frequently. Collecting
1108 pond. When Y merges into Z, things get epoch and the collecting pond
1109 gets smaller. What exactly are "missing files"?
1111 take note of current epoch of the alpha file, let's call it the
1114 find all files on disk
1116 remove all files registered in the recentworld up to recent-ts
1118 remove all files that have been deleted after recent-ts according to
1121 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1123 * rersyncrecent might be a cronjob with a (locked) state file which
1124 contains things like after and maybe last z sync or such?
1126 rrr-mirror might be an alternative name but how would we justify the
1127 three Rs when there is no Re-Rsync-Recent?
1129 With the --loop parameter it is an endless loop, without it is no loop.
1130 At least this is simple.
1132 * todo: new accssor z-interval specifies how often the Z file is updated
1133 against the filesystem. We probably want no epoch stamp on these
1134 entries. And we want to be able to filter the entries (e.g. no
1135 by-modules and by-category tree)
1137 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1139 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1140 into Z. We do not need the other files filled up with filesystem
1143 * need interface to query for a file in order to NOT call update on
1144 PAUSE a second time within a short time.
1146 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1148 * recommended update interval? Makes no sense, is different for
1155 change-log-default-name: "Todo"