1 2009-03-21 Andreas J. Koenig <andk@cpan.org>
3 * bug with native integers:
7 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
11 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
14 Native integer broke when native math was turned off. FIXED
16 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
17 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
20 Yes, records out of order:
23 447004 epoch: 995885533
24 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
27 447008 epoch: 995890358
28 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
31 447012 epoch: 995892221
32 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
37 * Bug: want the index files in a .recent directory
39 * Bug: lots of dot files are not deleted in time
41 * possible test case: can a delete change the timestamp? This would
42 probably break the order of events.
44 2009-03-20 Andreas J. Koenig <andk@cpan.org>
46 * 1233701831.34486 what's so special about this number/string? It
48 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
49 atm lives in Y,Q, and Z.
51 It is the first entry after id/--skip-locking which has timestamp
52 1234164228.11325 which represents a file that doesn't exist anymore.
56 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
57 _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
58 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
59 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
60 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
61 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
62 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
63 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
64 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
65 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
66 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
67 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
68 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
70 FIXED, it was the "--skip-locking" file where manual intervention was participating
72 * bug on the mirroring slave: when the dirtymark gets increased we
73 probably do not reset the done intervals. The mirrorer stays within
74 tight bounds where it tries to sync with upstream and never seems to
75 finish. In the debugging state file I see lots of identical intervals
76 that do not get collapsed. When I restart the mirrorer it dies with:
78 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
79 _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
80 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
81 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
82 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
83 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
84 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
85 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
86 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
87 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
89 and debugging stands at
95 epoch: 1237400802.5789
96 path: id/J/JO/JOHND/CHECKSUMS
99 epoch: 1237400807.97514
100 path: id/M/MI/MIYAGAWA/CHECKSUMS
103 epoch: 1237400817.94363
104 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
112 and it is reproducable.
114 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
115 pause has a fresh one.
117 FIXED, it was the third anded term in each of the ifs in the IV block in
118 _register_one: with that we make sure that we do not stamp on valuable
121 2009-03-17 Andreas J. Koenig <andk@cpan.org>
123 * done: verified the existence of the floating point bug in bleadperl
124 and verified that switching from YAML::Syck to YAML::XS does not resolve
127 BTW, the switch was doable with
129 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
131 and should be considered as a separate TODO
133 * todo: integrate a dirty update with two aggregate calls before
134 unlocking for frictionless dirtying
136 * todo: start the second rsync daemon on pause
138 * todo: move index files to .recent: this cannot simply be done by
139 setting filenameroot to .recent/RECENT. Other parts of the modules rely
140 on the fact that dirname(recentfile) is the root of the mirrored tree.
142 2009-03-16 Andreas J. Koenig <andk@cpan.org>
144 * What was the resolution of the mirror.pl delete hook bug? Do we call
145 the delete hook when pause removes a file from MUIR?
147 * Today on pause: Updating 2a13fba..29f284d and installing it for
148 /usr/local/perl-5.10.0{,-RC2}
150 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
151 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
153 2009-03-15 Andreas J. Koenig <andk@cpan.org>
155 * currently recent_events has the side effect of setting dirtymark
156 because it forces a read on the file. That should be transparent, so
157 that the dirtymark call always forces a cache-able(?) read.
159 * The bug below is -- after a lot of trying -- not reproducible on a
160 small script, only in the large test script. The closest to the output
166 my $x = "01237123229.8814";
171 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
173 $l = "1237123231.22458";
174 $r = "1237123231.22458";
176 Devel::Peek::Dump $l;
177 Devel::Peek::Dump $r;
178 Devel::Peek::Dump $x = $l <=> $r;
182 The checked in state at c404a85 fails the test with my
183 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
184 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
185 in the test script always consistent.
187 * Todo: write a test that inserts a second dirty file with an already
188 existing timestamp. DONE
190 * Bug in perl 5.10 on my 64bit box:
192 DB<98> Devel::Peek::Dump $l
193 SV = PVMG(0x19e0450) at 0x142a550
195 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
197 NV = 1237123231.22458
198 PV = 0x194ce70 "1237123231.22458"\0
202 DB<99> Devel::Peek::Dump $r
203 SV = PVMG(0x19e0240) at 0x142a3e8
205 FLAGS = (PADMY,POK,pPOK)
208 PV = 0x19ff900 "1237123231.22458"\0
212 DB<100> Devel::Peek::Dump $l <=> $r
213 SV = IV(0x19ea6e8) at 0x19ea6f0
215 FLAGS = (PADTMP,IOK,pIOK)
218 DB<101> Devel::Peek::Dump $l
219 SV = PVMG(0x19e0450) at 0x142a550
221 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
223 NV = 1237123231.22458
224 PV = 0x194ce70 "1237123231.22458"\0
228 DB<102> Devel::Peek::Dump $r
229 SV = PVMG(0x19e0240) at 0x142a3e8
231 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
233 NV = 1237123231.22458
234 PV = 0x19ff900 "1237123231.22458"\0
238 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
240 * Todo: reset "done" or "covered" and "minmax" after a dirty operation?
242 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
244 * $obj->merge ($other) needs to learn about equal epoch which may happen
245 since dirty_epoch intruded.
247 * Wontfix anytime soon: I think we currently do not support mkdir. Only
250 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
252 * Todo: continue working on update(...,$dirty_epoch). It must be
253 followed by a fast_aggregate!
255 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
257 * maybe we need a closest_entry or fitting_interval or something like
258 that. We want to merge an event into the middle of some recentfile.
259 First we do not know which file, then we do not know where to lock,
260 where to enter the new item, when and where to correct the dirtymark.
262 So my thought is we should first find which file.
264 Another part of my brain answers: what would happen if we would enter
265 the new file into the smallest file just like an ordinary new event,
266 just as an old event?
268 (1) we would write a duplicate timestamp? No, this would be easy to
271 (2) we would make the file large quickly? Yes, but so what? We are
272 changing the dirtymark, so are willing to disturb the downstream hosts.
274 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
276 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
280 https://rt.cpan.org/Ticket/Display.html?id=41199
282 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
284 Strangest output being HAKANARDO who managed to upload
286 Here is a better oneliner that includes also the first line of each
289 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
291 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
293 * memory leak in the syncher? It currently weighs 100M.
297 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
300 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
302 * lookup by epoch and by path and use this ability on the pause to never
303 again register a file twice that doesn't need it. Let's call it
306 * after the dirtymark is done: fill up recentfiles with fake (historic)
307 entries; fill up with individual corrections; algorithm maybe to be done
308 with bigfloat so that we can always place something in the middle
309 between two entries. Before we must switch to bigfloat we could try to
310 use Data::Float::nextup to get the.
312 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
315 * dirtymark now lives in Recentfile, needs to be used in rmirror.
317 * find out why the downloader died after a couple of hours without a net
318 connection. Write a test that survives the not-existence of the other
321 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
323 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
324 it makes less sense with tight coupling for secondary files. But asking
325 the server what the current epoch is might be cheaper on HTTP than on
326 rsync. (Needs to be evaluated)
328 * remove the 0.00 from the verbose overview in the Merged column in the
331 * write tests that expose the problems of the last few days: cascading
332 client/server roles, tight coupling for secondary RFs, deletes after
335 * Some day we might want to have policy options for the slave:
336 tight/loose/no coupling with upstream for secondary RFs. tight is what
337 we have now. loose would wait until a gap occurs that can be closed.
339 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
341 * revisit all $rfs->[$i+1] places if they now make sense still
343 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
345 * another bug is the fact that the mirror command deletes files before
346 it unhides the index file, thus confusing downstream slaves. We must not
347 delete before unhiding and must delete after unhiding. FIXED.
349 * new complication about the slave that is playing a server role.
350 Currently we mirror from newest to oldest with a hidden temporary file
351 as index. And when one file is finished, we unhide the index file.
352 Imagine the cascading server/slave is dead for a day. It then starts
353 mirroring again with the freshest thing and unhides the freshest index
354 file when it has worked through it. In that moment it exposes a time
355 hole. Because it now works on the second recentfile which is still
358 We currently do nothing special to converge after such a drop out. At
359 least not intentionally and robustly and thought through.
361 The algorithm we use to seed the next file needs quite a lot of more
362 robustness than it currently has. Something to do with looking at the
363 merged element of the next rf and when it has dropped off, we seed
364 immediately. And if it ramains dropped off, we seed again, of course.
366 Nope, looking from smaller to larger RFS we look at the merged element
367 of this RF and at the minmax/max element of the next RF. If that
368 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
371 And we need a public accessor seed and unseed or seeded. But not the mix
372 of public and private stuff that then is used behind the back.
374 And then the secondary* stuff must go.
376 And we must understand what the impact is on the DONE system. Can it go
377 unnoticed that there was a hole? And could the DONE system have decided
378 the hole is covered? This should be testable with three directories where
379 the middle stops working for a while. Done->merge is suspicious, we must
380 stop it from merging non-conflatable neighbors due to broken continuity.
384 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
386 * Slaven suggests to have the current epoch or the whole current
387 recentfile available from the HTTP server and take it away with
388 keepalive. This direction goes the granularity down to subseconds.
390 We might want to rewrite everything to factor out transport and allow
391 the whole thing to run via HTTP.
393 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
395 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
398 * smoker on k81 fetching from k75 to verify cascading works. See
399 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
401 * maybe the loop should wait for CHECKSUMS file after every upload. And
402 CPAN.pm needs to deal with timestamps in the future.
404 * do not forget the dirtymark!
406 Text: have a new flag on recentfiles with the meaning: if this
407 changes, you're required to run a full rsync over all the files. The
408 reason why we set it would probably be: some foul happened. we injected
409 files in arbitrary places or didn't inject them although they changed.
410 The content of the flag? Timestamp? The relation between the
411 recentfiles would have to be inheritance from the principal, because any
412 out of band changes would soon later propagate to the next recentfile.
414 By upping the flag often one can easily ruin the slaves.
416 last out of band change? dirtymark?
418 Anyway, this implies that we read a potentially existing recentfile
421 And it implies that we have an eventloop that keeps us busy in 2-3
422 cycles, one for current stuff (tight loop) and one for the recentfiles
423 (cascade when principal has changed), one for the old stuff after a
426 And it implies that the out-of-band change in any of the recentfiles
427 must have a lock on the principal file and there is the place to set the
430 * start a FAQ, especially quick start guide questions. Also to aid those
431 problematic areas where we have no good solution, like the "links"
434 * wish feedback when we are slow.
438 * Remove a few DEBUG statements.
440 * The multiple-rrr way of doing things needs a new option to rmirror,
441 like piecemeal or so. Not urgent because after the first pass through,
442 things run smoothely. It's only ugly during the first pass.
444 * I have the suspicion that the code is broken that decides if the
445 neighboring RF needs to be seeded. I fear when too much time has gone
446 between two calls (in our case more than one hour), it would not seed
447 the neighbor. Of course this will never be noticed, so we need a good
450 * local/localroot confusion: I currently pass both options but one must
453 * accounts for early birds on PAUSE rsync daemon.
455 * hardcoded 20 seconds
457 * who mirrors the index? DOING now.
459 * which CPAN mirrors offer rsync?
461 * visit all XXX, visit all _float places
463 * rename the pathdb stuff, it's too confusing. No idea how.
465 * rrr-inotify, backpan, rrr-register
467 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
469 * current bugs: the pathdb seems to get no reset, the seeding of the
470 secondaryttl stuff seems not to have an effect. Have helped myself with
471 a rand(10), need to fix this back. So not checked in. Does the rand
474 The rand thing helps. The secondaryttl stuff was in the wrong line,
477 The pathdb stuff was because I called either _pathdb or __pathdb on the
478 wrong object. FIXED now.
480 * It's not so beautiful if we never fetch the recentfiles that are not
481 the principal, even if this is correct behaviour. We really do not need
482 them after we have fetched the whole content.
484 OK, we want a switch for that: secondaryttl DONE
486 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
488 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
489 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
490 it lets 40 follow. FIXED
492 * See that the long running process really only updates the principal
493 file unless it has missed a timespan during which something happened. If
494 nothing happened, it must notice even when it misses the timespan. DONE
496 * we must throw away the pathdb when we have reached the end of Z. From
497 that moment we can have a very small pathdb because the only reason for
498 a pathdb is that we know to ignore old records in old files. We won't
499 need this pathdb again before the next full pass over the data is
500 necessary and then we will rebuild it as we go along. DONE
502 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
504 * I think, Done::register_one is doing wrong in that it does not
505 conflate neighboring pieces. The covered() method cannot do this because
506 it has no recent_events array at hand. But register_one has it and could
507 do it and for some reason misses to do it (sometimes).
509 This means that the three tests I just wrote can probably not survive
510 because they test with an already broken Done structure.
512 The art now is to detect how it happens, then to reproduce, then write a
515 So from the logfile this is what happens: we have a good interval with
516 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
517 top of it. Locally we now mirror F2 and open a new done interval for it.
518 Then we mirror F1 but this time with the timestamp T1b. And when we then
519 try to close the gap, we do not find T1 but instead something older. We
520 should gladly accept this older piece and this would fix this bug.
524 * bug to fix: when the 1h file changes while rmirror is running, we do
525 correctly sync the new files but never switch to the 6h file but rather
526 stay in a rather quick loop that fetches the 1h file again and again.
528 Is it possible that we initialize a new object? Or does
529 get_remote_recentfile_as_tempfile overwrite something in myself?
531 Want a new option: _runstatusfile => $file which frequently dumps the
532 state of all recentfiles to a file.
536 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
538 * Todo: now teach update to verify the timestamp is about to write
539 against the previous and use _increase_a_bit if it doesn't comply with
540 strict monotony. DONE
542 * The problem of rounding. So far perl's default precision was
543 sufficient. One day it won't be. FakeFloat has an easy job when it is
544 only reading and other machines have written correctly. But when we want
545 to write a floating point number that is a bit larger than the other
546 one, then we need our own idea of precision.
548 Slaven said: just append a "1". This might be going towards the end of
549 usability too quickly. I'd like something that actually uses the decimal
550 system. Well, appending a 1 also does this but...
552 E.g. we have 1.0. nextup on this architecture is starting with
553 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
554 taken the 1.0000000000000003 and the next user comes and the time tells
555 him 1.0 again. He has to beat my number without stepping over the
556 nextup. This is much less space than I had when I chose 1,2,3.
558 What is also irritating is that nextup is architecture dependent. The
559 128 bit guy must choose very long numbers to fit in between whereas the
560 other one with 16 bit uses larger steps. But then the algorithm is the
561 same for both, so that would be a nice thing.
563 I see two situation where we need this. One is when Time::HiRes returns
564 us a value that is <= the last entry in our recentfile. In this case
565 (let's call it the end-case) we must fill the region between that number
566 and the next higher native floating point number. The other is when we
567 inject an old file into an old recentfile (we would then also set a new
568 dirtymark). We find the integer value already taken and need a slightly
569 different one (let's call it the middle-case). The difference between
570 the two situations is that the next user will want to find something
571 higher than my number in the end-case and something lower than my number
574 So I suggest we give the function both a value and an upper bound and it
575 calculates us a primitive middle. The upper bound in the middle-case is
576 the next integer. The upper bound on the end-case is the nextup floating
577 point number. But the latter poses another problem: if we have occupied
578 the middle m between x and nextup(x), then the nextup(m) will probably
579 not be the same as nextup(x) because some rounding will take place
580 before the nextup is calculated and when the rounding reaches the
581 nextup(x), we will end up at nextup(nextup(x)).
583 So we really need to consider the nextup and the nextdown from there and
584 then the middle and that's the number we may approach asymptotically.
587 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
589 * consider deprecating the use of RECENT.recent as a symlink. It turns
590 out to need extra hoops with the rsync options and just isn't worth it.
591 Or maybe these extra hoops are needed anyway for the rest of the tree?
592 Nope, can't be the case because not all filesystems support symlinks.
594 But before doing the large step, I'll deprecate the call of
595 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
596 only introduced to resolve RECENT.recent and complicates the routine far
597 beyond what it deserves.
599 DONE. Won't deprecate RECENT.recent, just moved its handling to the
602 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
604 * I think it's a bug that the rsync_option links must be set to true in
605 order to support RECENT.recent and that nobody cares to set it
606 automatically. Similar for ignore_link_stat_errors.
608 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
610 * Todo: find all todos together and make a plan what is missing for a
613 - verifytree or something like that. fsck maybe.
615 - rersyncrecent, the script itself? What it do?
617 - a way to only mirror the recentfiles without mirroring the whole
618 remote system such that people can decide to mirror only partially see
619 also 2008-08-30. .shadow-xxx directory? this also needed for a
620 filesystem that is still incomplete and might need the mirrorfiles for
623 - long living objects that mirror again and again. Inject something
624 into ta, see how it goes over to tb.
626 - how do we continue filling up the DONE system when we use an object
627 for the second time? "fully covered" and "uptodate" or new terminology.
629 - overview called on the wrong file should be understandable
631 - the meta data field that must change when we fake something up so that
632 the downstream people know they have to re-fetch everything.
634 - how tolerant are we against missing files upstream? how do we keep
635 track? there are legitimate cases where we did read upstream index right
636 before a file got deleted there and then find that file as new and want
637 it. There are other cases that are not self healing and must be tracked
640 - how, exactly, do we have to deal with deletes? With rsync errors?
642 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
643 authors) failed: No such file or directory (2)
645 The file above is a delete in 1h and a new in file 1M and the
646 delete in the locally running rmirror did not get propagated to the 1M
647 object. Bug. And the consequence is a standstill.
649 It seems that a slave that works with a file below the principal needs
650 to merge things all the way up to get rid of later deletes. Or keep
651 track of all deletes and skip them later. So we need a trackdeletes.pm
652 similar to the done.pm?
654 see also 2008-08-20 about spurious deletes that really have no add
655 counterpart and yet they are not wrong.
657 - consider the effect when resyncing the recentfile takes longer than
658 the time per loop. Then we never rsync any file. We need to diagnose
659 that and force an increase of that loop time. But when we later are fast
660 enough again because the net has recovered, then we need to switch back
661 to original parameters. ERm, no, it's enough to keep syncing at least
662 one file before refetching an index file.
664 - remember to verify that no temp files are left lying around and the
667 - status file for not long running jobs that want to track upstream with
670 - revisit all XXX _float areas and study Sub::Exporter DONE
672 - persistent DB even though we just said we do not need it. Just for
673 extended capabilities and time savings when, for example, upstream
674 announces a reset and we get new recentfiles and could then limit
675 ourselves to a subset of files (those that have a changed epoch) in a
676 first pass and would only then do the loop to verify the rest. Or
679 * Todo: aggregate files should know their feed and finding the principal
680 should be done stepwise. (?)
682 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
683 left locks around. DONE
685 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
687 * maybe extend the _overview so that it always says if and where the
688 last file is in the next file and where the next event in the next rf
689 would lie. No, don't like this anymore. REJECT
691 * take the two new redundant tests out again, only the third must
694 * Todo: add a sanity check if the merged structure is really pointing to
695 a different rf and that this different rf is larger. DONE
697 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
699 * now test, if they are overlapping. And test if there is a file in the
700 next rf that would fit into this rf's interval.
702 1h 1222324012.8474 1222322541.7963 0.4086
703 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
704 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
705 1W 1222313218.3626 1221708477.5829 167.9835
707 I suspect that somebody writes a merged timestamp without having merged
708 and then somebody else relies on it.
710 If aggregate is running, the intervals must not be extravagated, if it
711 is not running, there must not be bounds, the total number of events in
712 the system must be counted and must be controlled throughout the tests.
713 That the test required the additional update was probably nonsense,
714 because aggregate can cut pieces too. FIXED & DONE
716 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
718 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
721 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
723 * Most apparent bug at the moment is that the recentfiles are fetched
724 too often. Only the principal should be fetched and if it has not
725 changed, the others should not be refetched. ATM I must admit that I'm
726 happy that we refetch more often than needed because I can more easily
727 fix bugs while the thing is running.
729 * Let's say, 1220474966.19501 is a timestamp of a file that is already
730 done but the done system does not know about it. The reason for the
731 failure is not known and we never reach the status uptodate because of
732 this. We must get over it.
734 Later it turns out that the origin server had a bug somewhere.
735 1220474966.19042 came after 1220474966.19501. Or better: it was in the
736 array of the recentfile one position above. The bug was my own.
738 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
740 * There is the race condition where the server does a delete and the
741 slave does not yet know and then tries to download it because he sees
742 the new. So for this time window we must be more tolerant against
743 failure. If we cannot download a file, we should just skip it and should
744 not retry immediately. The whole system should discover the lost thing
745 later. Keeping track with the DONE system should really be a no brainer.
747 But there is something more: the whole filesystem is a database and the
748 recentfiles are one possible representation of it. It's a pretty useful
749 representation I think that's why I have implemented something around
750 it. But for strictly local operation it has little value. For local
751 operation we would much rather have a database. So we would enter every
752 recentfile reading and every rsync operation and for every file the last
753 state change and what it leads to. Then we would always ignore older
754 records without the efforts involved with recentfiles.
756 The database would have: path,recentepoch,rsyncedon,deletedon
758 Oh well, not yet clear where this leads to.
760 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
762 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
763 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
765 * statusfile, maybe only on demand, alone to have a sharp debugging
766 tool. It is locked and all recentfiles dump themselves into it and we
767 can build a viewer that lets us know where we stand and what's inside.
769 * remember: only the principal recentfile needs expiration, all others
770 shall be expired by principal if it discovers that something has move
773 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
775 * Always check if we stringify to a higher value than in the entry
778 * And in covered make an additional check if we would be able to see a
779 numerical difference between the two numbers and if we can't then switch
780 to a different, more expensive algorithm. Do not want to be caught by
781 floating surprises. DONE
783 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
785 * caching has several aspects here: we can cache the interval of the
786 recentfile which only will change when the mtime of the file changes. We
787 must re-mirror the recentfile when its ttl has expired. Does have_read
788 tell you anything? It counts nothing at all. Only the mtime is
789 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
790 thing the have_mirrored because it is unrelated to the mtime.
792 * Robustness of floating point calculations! I always thought that the
793 string calculated by the origin server for the floating representation
794 of the epoch time is just a string. When we convert it to a number and
795 later back to a string, the other computer might come to a different
796 conclusion. This must not happen, we want to preserve it under any
797 circumstances. I will have to write tests with overlong sequences that
798 get lost in arithmetic and must see if all still works well. DONE
800 But one fragile point remains: if one host considers a>b and the other
801 one considers them == but no eq. To prevent this, we must probably do
802 some extra homework. DONE
804 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
806 * the concept of tracking DONE needs an object per recentfile that has
807 something like these methods:
809 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
811 covered() register() covered()
813 The unclear thing is how we translate points in time into intervals. We
814 could pass a reference to the current recent_events array when running
815 we_have(xxx) and let the DONE object iterate over it such that it only
816 has to store a list of intervals that can melt into each other. Ah, even
817 passing the list together with a list of indexes seems feasiable.
819 Or maybe ask for the inverted list?
821 Whenever the complete array is covered by the interval we say we are
822 fully covered and if the recentfile is not expired, we are uptodate.
824 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
826 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
828 * need a way to "return" the next entry after the end of a list. When
829 the caller says "before" or "after" we would like to know if he could
830 cover that interval/threshold or not because this influences the effect
831 of a newer timestamp of that recentfile. DONE with $opt{info}.
833 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
835 * one of the next things to tackle: the equivalent of csync2 -TIXU.
837 loop implies tixu (?). Nope, something like --statefile decides. Per
840 T test, I init, X including removals, U nodirtymark
842 So we have no concept of dirtymarks, we only trust that since we are
843 running we have observed everything steadily. But people will not let
844 this program run forever so we must consider both startup penalty and
845 book keeping for later runs. We keep this for later. For now we write a
846 long running mirror that merges several intervals.
848 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
850 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
853 * rersyncrecent, the script: default to one week. The name of the switch
854 is --after. Other switches? --loop!
856 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
858 * need a switch --skip-deletes (?)
860 * need a switch --enduser that tells us that the whole tempfile
861 discipline is not needed when there is no downstream user. (?)
863 Without this switch we cannot have a reasonable recent.pl that just
864 displays the recent additions. Either we accept to download everything.
865 Or we download temporary files without the typical rsync protocol
868 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
869 File::Temp::tempdir, this might be a win.
871 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
873 * apropos missing: we have no push, we never know the downstream
874 servers. People who know their downstream hosts and want to ascertain
875 something will want additional methods we have never thought about, like
876 update or delete a certain file.
878 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
880 * tempted to refactor rmirror into resolve_symlink, localize, etc.
881 Curious if rsync_options=links equal 0 vs. 1 will make the expected
884 * rsync options: it's a bit of a pain that we usually need several rsync
885 options, like compress, links, times, checksum and that there is no
886 reasonable default except the original rsync default. I think wee can
887 safely assume that the rsync options are shared between all recentfile
888 instances within one recent tree.
890 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
892 * deletes: if a delete follows an add quickly enough it may happen that
893 a downstream mirror did not see the add at all! It seems this needs to
894 be mentioned somewhere. The point here is that even if the downstream is
895 never missing the principal timeframe it may encounter a "delete" that
896 has no complimentary "add" anywhere.
898 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
900 * I suspect the treat of metadata is incorrect during read or something.
901 The bug that I am watching is that between 06:08 and 06:09 the 6h file
902 contained more than 6 hours worth of data. At 06:08 we merged into the
903 1d file. We need to take snapshots of the 6h file over the course of an
904 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
907 Much worse: watching the 1h file: right at the moment (at 06:35) it
908 covers 1218867584-1219120397 which is 70 hours.
910 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
911 08:19:44 2008, that is when I checked out last time, so it seems to be
912 aggregating and never truncating?
914 No, correct is: it is never truncating; but wrong is: it is aggregating.
915 It does receive a lot of events from time to time from a larger file.
916 Somehow a large file gets merged into the small one and because the
917 "meta/merged" attribute is missing, nobody is paying attention. I
918 believe that I can fix this by making sure that metadata are honoured
919 during read. DONE and test adjusted.
921 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
923 * grand renaming plan
925 remotebase => remoteroot to fit well with localroot DONE
926 local_path() => localroot seems to me should already work DONE
927 recentfile_basename => rfilename no need to stress it has no slash DONE
929 filenameroot??? Doesn't seem too bad to me today. Maybe something like
930 kern? It would anyway need a deprecation cycle because it is an
931 important constructor.
933 * I like the portability that Data::Serializer brings us but the price
934 is that some day we might find out that it is slowing us a bit. We'll
937 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
939 * should we not enter the interval of the principal (or the interval of
940 the merging file?) in every aggregated/merged file?
942 * we should aim at a first release and give up on thinking about
943 sanitizing stuff and zloop. Let's just admit that a full traditional
944 rsync is the only available sanitizer ATM. Otherwise it's complicated
945 stuff: sanitizing on the origin server, sanitizing on the slaves,
946 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
947 get the basics out before this becomes a major cause for mess.
949 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
951 * On OSes not supporting symlinks we expect that RECENT.recent contains
952 the contents of the principal recentfile. Actually this is identical on
953 systems supporting symlinks. Simple, what follows from that is that we
954 need to keep the serializer in the metadata because we cannot read it
955 from the filename, doesn't it? Of course not. It's a chicken and egg
956 problem. This leaves us with the problem to actually parse the
957 serialized data to find out in which format it is. So who can do the 4
958 or 5 magics we wanted to support? File::LibMagic?
960 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
962 * remotebase and recentfile_basename are ugly names. Now that we need a
963 word for the shortest/principal/driving recentfile too we should do
966 localroot is good. rfile is good. local_path() is bad, local_path($path)
967 is medium, filenameroot() is bad, remotebase is bad, recentfile is
970 Up to now remotebase was the string that described the remote root
971 directory in rsync notation, like pause.perl.org::authors. And
972 recentfile_basename was "RECENT-1h.yaml".
974 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
976 * The test that was added in today's checkin is a good start for a test
977 of rmirror. We should have more methods in Recent.pm: verify,
978 addmissingfiles. We should verify the current tree, then rmirror it and
979 then verifytree the copy. We could then add some arbitrary file and let
980 it be discovered by addmissingfiles, then rmirror again and then
981 verifytree the copy again.
983 Then we could start stealing from csync2 sqlite database [no port to
984 OSX!] and fill a local DB. And methods to compare the database with the
985 recentfiles. Our strength is that in principle we could maintain state
986 with a single float. We have synced up to 1234567890.123456. If the Z
987 file does not add new files all we have to do is mirror the new ones and
990 This makes it clear that we should extend current protocol and declare
991 that we cheat when we add files too late, just to help the other end
992 keeping track. Ah yes, that's what was meant when zloop was mentioned
995 Maybe need to revisit File::Mirror to help me with this task.
997 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
999 * There must be an allow-me-to-truncate flag in every recentfile.
1000 Without it one could construct a sequence of updates winning the locking
1001 battle against the aggregator. Only if an aggregator has managed to
1002 merge data over to the next level, truncating can be allowed. DONE with
1005 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1007 * We should probably guarantee that no duplicates enter the aggregator
1010 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1012 * To get merge operation faster would need a good benchmark test. What
1013 02 spits out isn't reliable enough and is dominated by many other
1016 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1017 Author: Andreas J. Koenig <k@k75.(none)>
1018 Date: Sat Aug 2 07:58:04 2008 +0200
1022 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1023 Author: Andreas J. Koenig <k@k75.(none)>
1024 Date: Sat Aug 2 11:40:29 2008 +0200
1026 there was a lot of trying but the effect is hardly measurable with
1029 * overhead of connecting seems high. When setting
1030 max_files_per_connection to 1 we see that.
1032 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1034 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1036 25138.2215600014/3600 = 6.98283932222261
1038 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1039 be a bug in the aggregator. FIXED
1041 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1043 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1044 we have it, can we mirror it, mirror it!
1046 I fear this needs a new class which might be called
1047 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1048 and treat them as an entity. I realize that a single recentfile may be
1049 sufficient for certain tasks and that it is handy for the low level
1050 programmer but it is not nice to use. If there is a delete in the 1h
1051 file then the 6h file still contains it. Seekers of the best information
1052 need to combine at least some of the recentfiles most of the time.
1054 There is the place for the Z loop!
1056 But the combination is something to collect in a database, isn't it. Did
1057 csync2 just harrumph?
1059 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1061 * it just occurred to me that hosts in the same mirroring pool could
1062 help out each other even without rewriting the recentfile. Just fetch
1063 the stuff to mirror from several places, bingo. But that's something
1064 that should rather live in a separate package or in rsync directly.
1066 * cronjobs are unsuited because with ntp they would all come at the full
1067 minute and disturb each other. Besides that I'd hate to have a backbone
1068 with more than a few seconds latency.
1070 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1072 * a second rsync server with access control for PAUSE. Port? 873 is the
1073 standard port, let's take 8873.
1075 * if there were a filesystem based on this, it would have a slow access
1076 to inexistent files. It would probably provide wrong readdir (only based
1077 on current content) or also a slow one (based on a recentfile written
1078 after the call). But it would provide fast access to existing files. Or
1079 one would deliberately allow slightly blurred answers based on some
1080 sqlite reflection of the recentfiles.
1082 * todo: write a variant of mirror() that combines two or more
1083 recentfiles and treats them like one
1085 * todo: signal handler to remove the tempfile
1087 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1089 * now that we have the symlink I forgot how it should be used in
1092 * the z loop: add missing files to Z file. Just append them (instead of
1093 prepending). So one guy prepends something from the Y file from time to
1094 time and another guy appends something rather frequently. Collecting
1095 pond. When Y merges into Z, things get epoch and the collecting pond
1096 gets smaller. What exactly are "missing files"?
1098 take note of current epoch of the alpha file, let's call it the
1101 find all files on disk
1103 remove all files registered in the recentworld up to recent-ts
1105 remove all files that have been deleted after recent-ts according to
1108 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1110 * rersyncrecent might be a cronjob with a (locked) state file which
1111 contains things like after and maybe last z sync or such?
1113 rrr-mirror might be an alternative name but how would we justify the
1114 three Rs when there is no Re-Rsync-Recent?
1116 With the --loop parameter it is an endless loop, without it is no loop.
1117 At least this is simple.
1119 * todo: new accssor z-interval specifies how often the Z file is updated
1120 against the filesystem. We probably want no epoch stamp on these
1121 entries. And we want to be able to filter the entries (e.g. no
1122 by-modules and by-category tree)
1124 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1126 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1127 into Z. We do not need the other files filled up with filesystem
1130 * need interface to query for a file in order to NOT call update on
1131 PAUSE a second time within a short time.
1133 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1135 * recommended update interval? Makes no sense, is different for
1142 change-log-default-name: "Todo"