1 2009-03-21 Andreas J. Koenig <andk@cpan.org>
3 * possible test case: can a delete change the timestamp? This would
4 probably break the order of events.
6 2009-03-20 Andreas J. Koenig <andk@cpan.org>
8 * 1233701831.34486 what's so special about this number/string? It
10 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
11 atm lives in Y,Q, and Z.
13 It is the first entry after id/--skip-locking which has timestamp
14 1234164228.11325 which represents a file that doesn't exist anymore.
18 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
19 _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
20 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
21 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
22 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
23 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
24 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
25 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
26 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
27 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
28 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
29 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
30 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
32 FIXED, it was the "--skip-locking" file where manual intervention was participating
34 * bug on the mirroring slave: when the dirtymark gets increased we
35 probably do not reset the done intervals. The mirrorer stays within
36 tight bounds where it tries to sync with upstream and never seems to
37 finish. In the debugging state file I see lots of identical intervals
38 that do not get collapsed. When I restart the mirrorer it dies with:
40 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
41 _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
42 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
43 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
44 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
45 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
46 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
47 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
48 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
49 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
51 and debugging stands at
57 epoch: 1237400802.5789
58 path: id/J/JO/JOHND/CHECKSUMS
61 epoch: 1237400807.97514
62 path: id/M/MI/MIYAGAWA/CHECKSUMS
65 epoch: 1237400817.94363
66 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
74 and it is reproducable.
76 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
77 pause has a fresh one.
79 FIXED, it probably was the missing conflation of "done" intervals.
81 2009-03-17 Andreas J. Koenig <andk@cpan.org>
83 * done: verified the existence of the floating point bug in bleadperl
84 and verified that switching from YAML::Syck to YAML::XS does not resolve
87 BTW, the switch was doable with
89 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
91 and should be considered as a separate TODO
93 * todo: integrate a dirty update with two aggregate calls before
94 unlocking for frictionless dirtying
96 * todo: start the second rsync daemon on pause
98 * todo: move index files to .recent: this cannot simply be done by
99 setting filenameroot to .recent/RECENT. Other parts of the modules rely
100 on the fact that dirname(recentfile) is the root of the mirrored tree.
102 2009-03-16 Andreas J. Koenig <andk@cpan.org>
104 * What was the resolution of the mirror.pl delete hook bug? Do we call
105 the delete hook when pause removes a file from MUIR?
107 * Today on pause: Updating 2a13fba..29f284d and installing it for
108 /usr/local/perl-5.10.0{,-RC2}
110 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
111 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
113 2009-03-15 Andreas J. Koenig <andk@cpan.org>
115 * currently recent_events has the side effect of setting dirtymark
116 because it forces a read on the file. That should be transparent, so
117 that the dirtymark call always forces a cache-able(?) read.
119 * The bug below is -- after a lot of trying -- not reproducible on a
120 small script, only in the large test script. The closest to the output
126 my $x = "01237123229.8814";
131 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
133 $l = "1237123231.22458";
134 $r = "1237123231.22458";
136 Devel::Peek::Dump $l;
137 Devel::Peek::Dump $r;
138 Devel::Peek::Dump $x = $l <=> $r;
142 The checked in state at c404a85 fails the test with my
143 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
144 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
145 in the test script always consistent.
147 * Todo: write a test that inserts a second dirty file with an already
148 existing timestamp. DONE
150 * Bug in perl 5.10 on my 64bit box:
152 DB<98> Devel::Peek::Dump $l
153 SV = PVMG(0x19e0450) at 0x142a550
155 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
157 NV = 1237123231.22458
158 PV = 0x194ce70 "1237123231.22458"\0
162 DB<99> Devel::Peek::Dump $r
163 SV = PVMG(0x19e0240) at 0x142a3e8
165 FLAGS = (PADMY,POK,pPOK)
168 PV = 0x19ff900 "1237123231.22458"\0
172 DB<100> Devel::Peek::Dump $l <=> $r
173 SV = IV(0x19ea6e8) at 0x19ea6f0
175 FLAGS = (PADTMP,IOK,pIOK)
178 DB<101> Devel::Peek::Dump $l
179 SV = PVMG(0x19e0450) at 0x142a550
181 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
183 NV = 1237123231.22458
184 PV = 0x194ce70 "1237123231.22458"\0
188 DB<102> Devel::Peek::Dump $r
189 SV = PVMG(0x19e0240) at 0x142a3e8
191 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
193 NV = 1237123231.22458
194 PV = 0x19ff900 "1237123231.22458"\0
198 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
200 * Todo: reset "done" or "covered" and "minmax" after a dirty operation?
202 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
204 * $obj->merge ($other) needs to learn about equal epoch which may happen
205 since dirty_epoch intruded.
207 * Wontfix anytime soon: I think we currently do not support mkdir. Only
210 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
212 * Todo: continue working on update(...,$dirty_epoch). It must be
213 followed by a fast_aggregate!
215 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
217 * maybe we need a closest_entry or fitting_interval or something like
218 that. We want to merge an event into the middle of some recentfile.
219 First we do not know which file, then we do not know where to lock,
220 where to enter the new item, when and where to correct the dirtymark.
222 So my thought is we should first find which file.
224 Another part of my brain answers: what would happen if we would enter
225 the new file into the smallest file just like an ordinary new event,
226 just as an old event?
228 (1) we would write a duplicate timestamp? No, this would be easy to
231 (2) we would make the file large quickly? Yes, but so what? We are
232 changing the dirtymark, so are willing to disturb the downstream hosts.
234 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
236 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
240 https://rt.cpan.org/Ticket/Display.html?id=41199
242 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
244 Strangest output being HAKANARDO who managed to upload
246 Here is a better oneliner that includes also the first line of each
249 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
251 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
253 * memory leak in the syncher? It currently weighs 100M.
257 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
260 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
262 * lookup by epoch and by path and use this ability on the pause to never
263 again register a file twice that doesn't need it. Let's call it
266 * after the dirtymark is done: fill up recentfiles with fake (historic)
267 entries; fill up with individual corrections; algorithm maybe to be done
268 with bigfloat so that we can always place something in the middle
269 between two entries. Before we must switch to bigfloat we could try to
270 use Data::Float::nextup to get the.
272 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
275 * dirtymark now lives in Recentfile, needs to be used in rmirror.
277 * find out why the downloader died after a couple of hours without a net
278 connection. Write a test that survives the not-existence of the other
281 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
283 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
284 it makes less sense with tight coupling for secondary files. But asking
285 the server what the current epoch is might be cheaper on HTTP than on
286 rsync. (Needs to be evaluated)
288 * remove the 0.00 from the verbose overview in the Merged column in the
291 * write tests that expose the problems of the last few days: cascading
292 client/server roles, tight coupling for secondary RFs, deletes after
295 * Some day we might want to have policy options for the slave:
296 tight/loose/no coupling with upstream for secondary RFs. tight is what
297 we have now. loose would wait until a gap occurs that can be closed.
299 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
301 * revisit all $rfs->[$i+1] places if they now make sense still
303 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
305 * another bug is the fact that the mirror command deletes files before
306 it unhides the index file, thus confusing downstream slaves. We must not
307 delete before unhiding and must delete after unhiding. FIXED.
309 * new complication about the slave that is playing a server role.
310 Currently we mirror from newest to oldest with a hidden temporary file
311 as index. And when one file is finished, we unhide the index file.
312 Imagine the cascading server/slave is dead for a day. It then starts
313 mirroring again with the freshest thing and unhides the freshest index
314 file when it has worked through it. In that moment it exposes a time
315 hole. Because it now works on the second recentfile which is still
318 We currently do nothing special to converge after such a drop out. At
319 least not intentionally and robustly and thought through.
321 The algorithm we use to seed the next file needs quite a lot of more
322 robustness than it currently has. Something to do with looking at the
323 merged element of the next rf and when it has dropped off, we seed
324 immediately. And if it ramains dropped off, we seed again, of course.
326 Nope, looking from smaller to larger RFS we look at the merged element
327 of this RF and at the minmax/max element of the next RF. If that
328 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
331 And we need a public accessor seed and unseed or seeded. But not the mix
332 of public and private stuff that then is used behind the back.
334 And then the secondary* stuff must go.
336 And we must understand what the impact is on the DONE system. Can it go
337 unnoticed that there was a hole? And could the DONE system have decided
338 the hole is covered? This should be testable with three directories where
339 the middle stops working for a while. Done->merge is suspicious, we must
340 stop it from merging non-conflatable neighbors due to broken continuity.
344 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
346 * Slaven suggests to have the current epoch or the whole current
347 recentfile available from the HTTP server and take it away with
348 keepalive. This direction goes the granularity down to subseconds.
350 We might want to rewrite everything to factor out transport and allow
351 the whole thing to run via HTTP.
353 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
355 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
358 * smoker on k81 fetching from k75 to verify cascading works. See
359 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
361 * maybe the loop should wait for CHECKSUMS file after every upload. And
362 CPAN.pm needs to deal with timestamps in the future.
364 * do not forget the dirtymark!
366 Text: have a new flag on recentfiles with the meaning: if this
367 changes, you're required to run a full rsync over all the files. The
368 reason why we set it would probably be: some foul happened. we injected
369 files in arbitrary places or didn't inject them although they changed.
370 The content of the flag? Timestamp? The relation between the
371 recentfiles would have to be inheritance from the principal, because any
372 out of band changes would soon later propagate to the next recentfile.
374 By upping the flag often one can easily ruin the slaves.
376 last out of band change? dirtymark?
378 Anyway, this implies that we read a potentially existing recentfile
381 And it implies that we have an eventloop that keeps us busy in 2-3
382 cycles, one for current stuff (tight loop) and one for the recentfiles
383 (cascade when principal has changed), one for the old stuff after a
386 And it implies that the out-of-band change in any of the recentfiles
387 must have a lock on the principal file and there is the place to set the
390 * start a FAQ, especially quick start guide questions. Also to aid those
391 problematic areas where we have no good solution, like the "links"
394 * wish feedback when we are slow.
398 * Remove a few DEBUG statements.
400 * The multiple-rrr way of doing things needs a new option to rmirror,
401 like piecemeal or so. Not urgent because after the first pass through,
402 things run smoothely. It's only ugly during the first pass.
404 * I have the suspicion that the code is broken that decides if the
405 neighboring RF needs to be seeded. I fear when too much time has gone
406 between two calls (in our case more than one hour), it would not seed
407 the neighbor. Of course this will never be noticed, so we need a good
410 * local/localroot confusion: I currently pass both options but one must
413 * accounts for early birds on PAUSE rsync daemon.
415 * hardcoded 20 seconds
417 * who mirrors the index? DOING now.
419 * which CPAN mirrors offer rsync?
421 * visit all XXX, visit all _float places
423 * rename the pathdb stuff, it's too confusing. No idea how.
425 * rrr-inotify, backpan, rrr-register
427 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
429 * current bugs: the pathdb seems to get no reset, the seeding of the
430 secondaryttl stuff seems not to have an effect. Have helped myself with
431 a rand(10), need to fix this back. So not checked in. Does the rand
434 The rand thing helps. The secondaryttl stuff was in the wrong line,
437 The pathdb stuff was because I called either _pathdb or __pathdb on the
438 wrong object. FIXED now.
440 * It's not so beautiful if we never fetch the recentfiles that are not
441 the principal, even if this is correct behaviour. We really do not need
442 them after we have fetched the whole content.
444 OK, we want a switch for that: secondaryttl DONE
446 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
448 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
449 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
450 it lets 40 follow. FIXED
452 * See that the long running process really only updates the principal
453 file unless it has missed a timespan during which something happened. If
454 nothing happened, it must notice even when it misses the timespan. DONE
456 * we must throw away the pathdb when we have reached the end of Z. From
457 that moment we can have a very small pathdb because the only reason for
458 a pathdb is that we know to ignore old records in old files. We won't
459 need this pathdb again before the next full pass over the data is
460 necessary and then we will rebuild it as we go along. DONE
462 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
464 * I think, Done::register_one is doing wrong in that it does not
465 conflate neighboring pieces. The covered() method cannot do this because
466 it has no recent_events array at hand. But register_one has it and could
467 do it and for some reason misses to do it (sometimes).
469 This means that the three tests I just wrote can probably not survive
470 because they test with an already broken Done structure.
472 The art now is to detect how it happens, then to reproduce, then write a
475 So from the logfile this is what happens: we have a good interval with
476 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
477 top of it. Locally we now mirror F2 and open a new done interval for it.
478 Then we mirror F1 but this time with the timestamp T1b. And when we then
479 try to close the gap, we do not find T1 but instead something older. We
480 should gladly accept this older piece and this would fix this bug.
484 * bug to fix: when the 1h file changes while rmirror is running, we do
485 correctly sync the new files but never switch to the 6h file but rather
486 stay in a rather quick loop that fetches the 1h file again and again.
488 Is it possible that we initialize a new object? Or does
489 get_remote_recentfile_as_tempfile overwrite something in myself?
491 Want a new option: _runstatusfile => $file which frequently dumps the
492 state of all recentfiles to a file.
496 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
498 * Todo: now teach update to verify the timestamp is about to write
499 against the previous and use _increase_a_bit if it doesn't comply with
500 strict monotony. DONE
502 * The problem of rounding. So far perl's default precision was
503 sufficient. One day it won't be. FakeFloat has an easy job when it is
504 only reading and other machines have written correctly. But when we want
505 to write a floating point number that is a bit larger than the other
506 one, then we need our own idea of precision.
508 Slaven said: just append a "1". This might be going towards the end of
509 usability too quickly. I'd like something that actually uses the decimal
510 system. Well, appending a 1 also does this but...
512 E.g. we have 1.0. nextup on this architecture is starting with
513 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
514 taken the 1.0000000000000003 and the next user comes and the time tells
515 him 1.0 again. He has to beat my number without stepping over the
516 nextup. This is much less space than I had when I chose 1,2,3.
518 What is also irritating is that nextup is architecture dependent. The
519 128 bit guy must choose very long numbers to fit in between whereas the
520 other one with 16 bit uses larger steps. But then the algorithm is the
521 same for both, so that would be a nice thing.
523 I see two situation where we need this. One is when Time::HiRes returns
524 us a value that is <= the last entry in our recentfile. In this case
525 (let's call it the end-case) we must fill the region between that number
526 and the next higher native floating point number. The other is when we
527 inject an old file into an old recentfile (we would then also set a new
528 dirtymark). We find the integer value already taken and need a slightly
529 different one (let's call it the middle-case). The difference between
530 the two situations is that the next user will want to find something
531 higher than my number in the end-case and something lower than my number
534 So I suggest we give the function both a value and an upper bound and it
535 calculates us a primitive middle. The upper bound in the middle-case is
536 the next integer. The upper bound on the end-case is the nextup floating
537 point number. But the latter poses another problem: if we have occupied
538 the middle m between x and nextup(x), then the nextup(m) will probably
539 not be the same as nextup(x) because some rounding will take place
540 before the nextup is calculated and when the rounding reaches the
541 nextup(x), we will end up at nextup(nextup(x)).
543 So we really need to consider the nextup and the nextdown from there and
544 then the middle and that's the number we may approach asymptotically.
547 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
549 * consider deprecating the use of RECENT.recent as a symlink. It turns
550 out to need extra hoops with the rsync options and just isn't worth it.
551 Or maybe these extra hoops are needed anyway for the rest of the tree?
552 Nope, can't be the case because not all filesystems support symlinks.
554 But before doing the large step, I'll deprecate the call of
555 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
556 only introduced to resolve RECENT.recent and complicates the routine far
557 beyond what it deserves.
559 DONE. Won't deprecate RECENT.recent, just moved its handling to the
562 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
564 * I think it's a bug that the rsync_option links must be set to true in
565 order to support RECENT.recent and that nobody cares to set it
566 automatically. Similar for ignore_link_stat_errors.
568 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
570 * Todo: find all todos together and make a plan what is missing for a
573 - verifytree or something like that. fsck maybe.
575 - rersyncrecent, the script itself? What it do?
577 - a way to only mirror the recentfiles without mirroring the whole
578 remote system such that people can decide to mirror only partially see
579 also 2008-08-30. .shadow-xxx directory? this also needed for a
580 filesystem that is still incomplete and might need the mirrorfiles for
583 - long living objects that mirror again and again. Inject something
584 into ta, see how it goes over to tb.
586 - how do we continue filling up the DONE system when we use an object
587 for the second time? "fully covered" and "uptodate" or new terminology.
589 - overview called on the wrong file should be understandable
591 - the meta data field that must change when we fake something up so that
592 the downstream people know they have to re-fetch everything.
594 - how tolerant are we against missing files upstream? how do we keep
595 track? there are legitimate cases where we did read upstream index right
596 before a file got deleted there and then find that file as new and want
597 it. There are other cases that are not self healing and must be tracked
600 - how, exactly, do we have to deal with deletes? With rsync errors?
602 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
603 authors) failed: No such file or directory (2)
605 The file above is a delete in 1h and a new in file 1M and the
606 delete in the locally running rmirror did not get propagated to the 1M
607 object. Bug. And the consequence is a standstill.
609 It seems that a slave that works with a file below the principal needs
610 to merge things all the way up to get rid of later deletes. Or keep
611 track of all deletes and skip them later. So we need a trackdeletes.pm
612 similar to the done.pm?
614 see also 2008-08-20 about spurious deletes that really have no add
615 counterpart and yet they are not wrong.
617 - consider the effect when resyncing the recentfile takes longer than
618 the time per loop. Then we never rsync any file. We need to diagnose
619 that and force an increase of that loop time. But when we later are fast
620 enough again because the net has recovered, then we need to switch back
621 to original parameters. ERm, no, it's enough to keep syncing at least
622 one file before refetching an index file.
624 - remember to verify that no temp files are left lying around and the
627 - status file for not long running jobs that want to track upstream with
630 - revisit all XXX _float areas and study Sub::Exporter DONE
632 - persistent DB even though we just said we do not need it. Just for
633 extended capabilities and time savings when, for example, upstream
634 announces a reset and we get new recentfiles and could then limit
635 ourselves to a subset of files (those that have a changed epoch) in a
636 first pass and would only then do the loop to verify the rest. Or
639 * Todo: aggregate files should know their feed and finding the principal
640 should be done stepwise. (?)
642 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
643 left locks around. DONE
645 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
647 * maybe extend the _overview so that it always says if and where the
648 last file is in the next file and where the next event in the next rf
649 would lie. No, don't like this anymore. REJECT
651 * take the two new redundant tests out again, only the third must
654 * Todo: add a sanity check if the merged structure is really pointing to
655 a different rf and that this different rf is larger. DONE
657 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
659 * now test, if they are overlapping. And test if there is a file in the
660 next rf that would fit into this rf's interval.
662 1h 1222324012.8474 1222322541.7963 0.4086
663 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
664 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
665 1W 1222313218.3626 1221708477.5829 167.9835
667 I suspect that somebody writes a merged timestamp without having merged
668 and then somebody else relies on it.
670 If aggregate is running, the intervals must not be extravagated, if it
671 is not running, there must not be bounds, the total number of events in
672 the system must be counted and must be controlled throughout the tests.
673 That the test required the additional update was probably nonsense,
674 because aggregate can cut pieces too. FIXED & DONE
676 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
678 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
681 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
683 * Most apparent bug at the moment is that the recentfiles are fetched
684 too often. Only the principal should be fetched and if it has not
685 changed, the others should not be refetched. ATM I must admit that I'm
686 happy that we refetch more often than needed because I can more easily
687 fix bugs while the thing is running.
689 * Let's say, 1220474966.19501 is a timestamp of a file that is already
690 done but the done system does not know about it. The reason for the
691 failure is not known and we never reach the status uptodate because of
692 this. We must get over it.
694 Later it turns out that the origin server had a bug somewhere.
695 1220474966.19042 came after 1220474966.19501. Or better: it was in the
696 array of the recentfile one position above. The bug was my own.
698 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
700 * There is the race condition where the server does a delete and the
701 slave does not yet know and then tries to download it because he sees
702 the new. So for this time window we must be more tolerant against
703 failure. If we cannot download a file, we should just skip it and should
704 not retry immediately. The whole system should discover the lost thing
705 later. Keeping track with the DONE system should really be a no brainer.
707 But there is something more: the whole filesystem is a database and the
708 recentfiles are one possible representation of it. It's a pretty useful
709 representation I think that's why I have implemented something around
710 it. But for strictly local operation it has little value. For local
711 operation we would much rather have a database. So we would enter every
712 recentfile reading and every rsync operation and for every file the last
713 state change and what it leads to. Then we would always ignore older
714 records without the efforts involved with recentfiles.
716 The database would have: path,recentepoch,rsyncedon,deletedon
718 Oh well, not yet clear where this leads to.
720 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
722 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
723 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
725 * statusfile, maybe only on demand, alone to have a sharp debugging
726 tool. It is locked and all recentfiles dump themselves into it and we
727 can build a viewer that lets us know where we stand and what's inside.
729 * remember: only the principal recentfile needs expiration, all others
730 shall be expired by principal if it discovers that something has move
733 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
735 * Always check if we stringify to a higher value than in the entry
738 * And in covered make an additional check if we would be able to see a
739 numerical difference between the two numbers and if we can't then switch
740 to a different, more expensive algorithm. Do not want to be caught by
741 floating surprises. DONE
743 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
745 * caching has several aspects here: we can cache the interval of the
746 recentfile which only will change when the mtime of the file changes. We
747 must re-mirror the recentfile when its ttl has expired. Does have_read
748 tell you anything? It counts nothing at all. Only the mtime is
749 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
750 thing the have_mirrored because it is unrelated to the mtime.
752 * Robustness of floating point calculations! I always thought that the
753 string calculated by the origin server for the floating representation
754 of the epoch time is just a string. When we convert it to a number and
755 later back to a string, the other computer might come to a different
756 conclusion. This must not happen, we want to preserve it under any
757 circumstances. I will have to write tests with overlong sequences that
758 get lost in arithmetic and must see if all still works well. DONE
760 But one fragile point remains: if one host considers a>b and the other
761 one considers them == but no eq. To prevent this, we must probably do
762 some extra homework. DONE
764 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
766 * the concept of tracking DONE needs an object per recentfile that has
767 something like these methods:
769 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
771 covered() register() covered()
773 The unclear thing is how we translate points in time into intervals. We
774 could pass a reference to the current recent_events array when running
775 we_have(xxx) and let the DONE object iterate over it such that it only
776 has to store a list of intervals that can melt into each other. Ah, even
777 passing the list together with a list of indexes seems feasiable.
779 Or maybe ask for the inverted list?
781 Whenever the complete array is covered by the interval we say we are
782 fully covered and if the recentfile is not expired, we are uptodate.
784 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
786 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
788 * need a way to "return" the next entry after the end of a list. When
789 the caller says "before" or "after" we would like to know if he could
790 cover that interval/threshold or not because this influences the effect
791 of a newer timestamp of that recentfile. DONE with $opt{info}.
793 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
795 * one of the next things to tackle: the equivalent of csync2 -TIXU.
797 loop implies tixu (?). Nope, something like --statefile decides. Per
800 T test, I init, X including removals, U nodirtymark
802 So we have no concept of dirtymarks, we only trust that since we are
803 running we have observed everything steadily. But people will not let
804 this program run forever so we must consider both startup penalty and
805 book keeping for later runs. We keep this for later. For now we write a
806 long running mirror that merges several intervals.
808 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
810 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
813 * rersyncrecent, the script: default to one week. The name of the switch
814 is --after. Other switches? --loop!
816 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
818 * need a switch --skip-deletes (?)
820 * need a switch --enduser that tells us that the whole tempfile
821 discipline is not needed when there is no downstream user. (?)
823 Without this switch we cannot have a reasonable recent.pl that just
824 displays the recent additions. Either we accept to download everything.
825 Or we download temporary files without the typical rsync protocol
828 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
829 File::Temp::tempdir, this might be a win.
831 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
833 * apropos missing: we have no push, we never know the downstream
834 servers. People who know their downstream hosts and want to ascertain
835 something will want additional methods we have never thought about, like
836 update or delete a certain file.
838 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
840 * tempted to refactor rmirror into resolve_symlink, localize, etc.
841 Curious if rsync_options=links equal 0 vs. 1 will make the expected
844 * rsync options: it's a bit of a pain that we usually need several rsync
845 options, like compress, links, times, checksum and that there is no
846 reasonable default except the original rsync default. I think wee can
847 safely assume that the rsync options are shared between all recentfile
848 instances within one recent tree.
850 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
852 * deletes: if a delete follows an add quickly enough it may happen that
853 a downstream mirror did not see the add at all! It seems this needs to
854 be mentioned somewhere. The point here is that even if the downstream is
855 never missing the principal timeframe it may encounter a "delete" that
856 has no complimentary "add" anywhere.
858 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
860 * I suspect the treat of metadata is incorrect during read or something.
861 The bug that I am watching is that between 06:08 and 06:09 the 6h file
862 contained more than 6 hours worth of data. At 06:08 we merged into the
863 1d file. We need to take snapshots of the 6h file over the course of an
864 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
867 Much worse: watching the 1h file: right at the moment (at 06:35) it
868 covers 1218867584-1219120397 which is 70 hours.
870 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
871 08:19:44 2008, that is when I checked out last time, so it seems to be
872 aggregating and never truncating?
874 No, correct is: it is never truncating; but wrong is: it is aggregating.
875 It does receive a lot of events from time to time from a larger file.
876 Somehow a large file gets merged into the small one and because the
877 "meta/merged" attribute is missing, nobody is paying attention. I
878 believe that I can fix this by making sure that metadata are honoured
879 during read. DONE and test adjusted.
881 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
883 * grand renaming plan
885 remotebase => remoteroot to fit well with localroot DONE
886 local_path() => localroot seems to me should already work DONE
887 recentfile_basename => rfilename no need to stress it has no slash DONE
889 filenameroot??? Doesn't seem too bad to me today. Maybe something like
890 kern? It would anyway need a deprecation cycle because it is an
891 important constructor.
893 * I like the portability that Data::Serializer brings us but the price
894 is that some day we might find out that it is slowing us a bit. We'll
897 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
899 * should we not enter the interval of the principal (or the interval of
900 the merging file?) in every aggregated/merged file?
902 * we should aim at a first release and give up on thinking about
903 sanitizing stuff and zloop. Let's just admit that a full traditional
904 rsync is the only available sanitizer ATM. Otherwise it's complicated
905 stuff: sanitizing on the origin server, sanitizing on the slaves,
906 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
907 get the basics out before this becomes a major cause for mess.
909 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
911 * On OSes not supporting symlinks we expect that RECENT.recent contains
912 the contents of the principal recentfile. Actually this is identical on
913 systems supporting symlinks. Simple, what follows from that is that we
914 need to keep the serializer in the metadata because we cannot read it
915 from the filename, doesn't it? Of course not. It's a chicken and egg
916 problem. This leaves us with the problem to actually parse the
917 serialized data to find out in which format it is. So who can do the 4
918 or 5 magics we wanted to support? File::LibMagic?
920 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
922 * remotebase and recentfile_basename are ugly names. Now that we need a
923 word for the shortest/principal/driving recentfile too we should do
926 localroot is good. rfile is good. local_path() is bad, local_path($path)
927 is medium, filenameroot() is bad, remotebase is bad, recentfile is
930 Up to now remotebase was the string that described the remote root
931 directory in rsync notation, like pause.perl.org::authors. And
932 recentfile_basename was "RECENT-1h.yaml".
934 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
936 * The test that was added in today's checkin is a good start for a test
937 of rmirror. We should have more methods in Recent.pm: verify,
938 addmissingfiles. We should verify the current tree, then rmirror it and
939 then verifytree the copy. We could then add some arbitrary file and let
940 it be discovered by addmissingfiles, then rmirror again and then
941 verifytree the copy again.
943 Then we could start stealing from csync2 sqlite database [no port to
944 OSX!] and fill a local DB. And methods to compare the database with the
945 recentfiles. Our strength is that in principle we could maintain state
946 with a single float. We have synced up to 1234567890.123456. If the Z
947 file does not add new files all we have to do is mirror the new ones and
950 This makes it clear that we should extend current protocol and declare
951 that we cheat when we add files too late, just to help the other end
952 keeping track. Ah yes, that's what was meant when zloop was mentioned
955 Maybe need to revisit File::Mirror to help me with this task.
957 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
959 * There must be an allow-me-to-truncate flag in every recentfile.
960 Without it one could construct a sequence of updates winning the locking
961 battle against the aggregator. Only if an aggregator has managed to
962 merge data over to the next level, truncating can be allowed. DONE with
965 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
967 * We should probably guarantee that no duplicates enter the aggregator
970 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
972 * To get merge operation faster would need a good benchmark test. What
973 02 spits out isn't reliable enough and is dominated by many other
976 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
977 Author: Andreas J. Koenig <k@k75.(none)>
978 Date: Sat Aug 2 07:58:04 2008 +0200
982 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
983 Author: Andreas J. Koenig <k@k75.(none)>
984 Date: Sat Aug 2 11:40:29 2008 +0200
986 there was a lot of trying but the effect is hardly measurable with
989 * overhead of connecting seems high. When setting
990 max_files_per_connection to 1 we see that.
992 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
994 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
996 25138.2215600014/3600 = 6.98283932222261
998 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
999 be a bug in the aggregator. FIXED
1001 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1003 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1004 we have it, can we mirror it, mirror it!
1006 I fear this needs a new class which might be called
1007 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1008 and treat them as an entity. I realize that a single recentfile may be
1009 sufficient for certain tasks and that it is handy for the low level
1010 programmer but it is not nice to use. If there is a delete in the 1h
1011 file then the 6h file still contains it. Seekers of the best information
1012 need to combine at least some of the recentfiles most of the time.
1014 There is the place for the Z loop!
1016 But the combination is something to collect in a database, isn't it. Did
1017 csync2 just harrumph?
1019 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1021 * it just occurred to me that hosts in the same mirroring pool could
1022 help out each other even without rewriting the recentfile. Just fetch
1023 the stuff to mirror from several places, bingo. But that's something
1024 that should rather live in a separate package or in rsync directly.
1026 * cronjobs are unsuited because with ntp they would all come at the full
1027 minute and disturb each other. Besides that I'd hate to have a backbone
1028 with more than a few seconds latency.
1030 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1032 * a second rsync server with access control for PAUSE. Port? 873 is the
1033 standard port, let's take 8873.
1035 * if there were a filesystem based on this, it would have a slow access
1036 to inexistent files. It would probably provide wrong readdir (only based
1037 on current content) or also a slow one (based on a recentfile written
1038 after the call). But it would provide fast access to existing files. Or
1039 one would deliberately allow slightly blurred answers based on some
1040 sqlite reflection of the recentfiles.
1042 * todo: write a variant of mirror() that combines two or more
1043 recentfiles and treats them like one
1045 * todo: signal handler to remove the tempfile
1047 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1049 * now that we have the symlink I forgot how it should be used in
1052 * the z loop: add missing files to Z file. Just append them (instead of
1053 prepending). So one guy prepends something from the Y file from time to
1054 time and another guy appends something rather frequently. Collecting
1055 pond. When Y merges into Z, things get epoch and the collecting pond
1056 gets smaller. What exactly are "missing files"?
1058 take note of current epoch of the alpha file, let's call it the
1061 find all files on disk
1063 remove all files registered in the recentworld up to recent-ts
1065 remove all files that have been deleted after recent-ts according to
1068 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1070 * rersyncrecent might be a cronjob with a (locked) state file which
1071 contains things like after and maybe last z sync or such?
1073 rrr-mirror might be an alternative name but how would we justify the
1074 three Rs when there is no Re-Rsync-Recent?
1076 With the --loop parameter it is an endless loop, without it is no loop.
1077 At least this is simple.
1079 * todo: new accssor z-interval specifies how often the Z file is updated
1080 against the filesystem. We probably want no epoch stamp on these
1081 entries. And we want to be able to filter the entries (e.g. no
1082 by-modules and by-category tree)
1084 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1086 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1087 into Z. We do not need the other files filled up with filesystem
1090 * need interface to query for a file in order to NOT call update on
1091 PAUSE a second time within a short time.
1093 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1095 * recommended update interval? Makes no sense, is different for
1102 change-log-default-name: "Todo"