1 2009-03-21 Andreas J. Koenig <andk@cpan.org>
3 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
4 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
7 Yes, records out of order:
10 447004 epoch: 995885533
11 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
14 447008 epoch: 995890358
15 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
18 447012 epoch: 995892221
19 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
24 * Bug: want the index files in a .recent directory
26 * Bug: lots of dot files are not deleted in time
28 * possible test case: can a delete change the timestamp? This would
29 probably break the order of events.
31 2009-03-20 Andreas J. Koenig <andk@cpan.org>
33 * 1233701831.34486 what's so special about this number/string? It
35 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
36 atm lives in Y,Q, and Z.
38 It is the first entry after id/--skip-locking which has timestamp
39 1234164228.11325 which represents a file that doesn't exist anymore.
43 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
44 _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
45 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
46 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
47 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
48 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
49 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
50 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
51 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
52 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
53 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
54 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
55 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
57 FIXED, it was the "--skip-locking" file where manual intervention was participating
59 * bug on the mirroring slave: when the dirtymark gets increased we
60 probably do not reset the done intervals. The mirrorer stays within
61 tight bounds where it tries to sync with upstream and never seems to
62 finish. In the debugging state file I see lots of identical intervals
63 that do not get collapsed. When I restart the mirrorer it dies with:
65 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
66 _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
67 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
68 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
69 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
70 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
71 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
72 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
73 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
74 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
76 and debugging stands at
82 epoch: 1237400802.5789
83 path: id/J/JO/JOHND/CHECKSUMS
86 epoch: 1237400807.97514
87 path: id/M/MI/MIYAGAWA/CHECKSUMS
90 epoch: 1237400817.94363
91 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
99 and it is reproducable.
101 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
102 pause has a fresh one.
104 FIXED, it probably was the missing conflation of "done" intervals.
106 2009-03-17 Andreas J. Koenig <andk@cpan.org>
108 * done: verified the existence of the floating point bug in bleadperl
109 and verified that switching from YAML::Syck to YAML::XS does not resolve
112 BTW, the switch was doable with
114 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
116 and should be considered as a separate TODO
118 * todo: integrate a dirty update with two aggregate calls before
119 unlocking for frictionless dirtying
121 * todo: start the second rsync daemon on pause
123 * todo: move index files to .recent: this cannot simply be done by
124 setting filenameroot to .recent/RECENT. Other parts of the modules rely
125 on the fact that dirname(recentfile) is the root of the mirrored tree.
127 2009-03-16 Andreas J. Koenig <andk@cpan.org>
129 * What was the resolution of the mirror.pl delete hook bug? Do we call
130 the delete hook when pause removes a file from MUIR?
132 * Today on pause: Updating 2a13fba..29f284d and installing it for
133 /usr/local/perl-5.10.0{,-RC2}
135 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
136 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
138 2009-03-15 Andreas J. Koenig <andk@cpan.org>
140 * currently recent_events has the side effect of setting dirtymark
141 because it forces a read on the file. That should be transparent, so
142 that the dirtymark call always forces a cache-able(?) read.
144 * The bug below is -- after a lot of trying -- not reproducible on a
145 small script, only in the large test script. The closest to the output
151 my $x = "01237123229.8814";
156 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
158 $l = "1237123231.22458";
159 $r = "1237123231.22458";
161 Devel::Peek::Dump $l;
162 Devel::Peek::Dump $r;
163 Devel::Peek::Dump $x = $l <=> $r;
167 The checked in state at c404a85 fails the test with my
168 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
169 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
170 in the test script always consistent.
172 * Todo: write a test that inserts a second dirty file with an already
173 existing timestamp. DONE
175 * Bug in perl 5.10 on my 64bit box:
177 DB<98> Devel::Peek::Dump $l
178 SV = PVMG(0x19e0450) at 0x142a550
180 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
182 NV = 1237123231.22458
183 PV = 0x194ce70 "1237123231.22458"\0
187 DB<99> Devel::Peek::Dump $r
188 SV = PVMG(0x19e0240) at 0x142a3e8
190 FLAGS = (PADMY,POK,pPOK)
193 PV = 0x19ff900 "1237123231.22458"\0
197 DB<100> Devel::Peek::Dump $l <=> $r
198 SV = IV(0x19ea6e8) at 0x19ea6f0
200 FLAGS = (PADTMP,IOK,pIOK)
203 DB<101> Devel::Peek::Dump $l
204 SV = PVMG(0x19e0450) at 0x142a550
206 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
208 NV = 1237123231.22458
209 PV = 0x194ce70 "1237123231.22458"\0
213 DB<102> Devel::Peek::Dump $r
214 SV = PVMG(0x19e0240) at 0x142a3e8
216 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
218 NV = 1237123231.22458
219 PV = 0x19ff900 "1237123231.22458"\0
223 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
225 * Todo: reset "done" or "covered" and "minmax" after a dirty operation?
227 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
229 * $obj->merge ($other) needs to learn about equal epoch which may happen
230 since dirty_epoch intruded.
232 * Wontfix anytime soon: I think we currently do not support mkdir. Only
235 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
237 * Todo: continue working on update(...,$dirty_epoch). It must be
238 followed by a fast_aggregate!
240 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
242 * maybe we need a closest_entry or fitting_interval or something like
243 that. We want to merge an event into the middle of some recentfile.
244 First we do not know which file, then we do not know where to lock,
245 where to enter the new item, when and where to correct the dirtymark.
247 So my thought is we should first find which file.
249 Another part of my brain answers: what would happen if we would enter
250 the new file into the smallest file just like an ordinary new event,
251 just as an old event?
253 (1) we would write a duplicate timestamp? No, this would be easy to
256 (2) we would make the file large quickly? Yes, but so what? We are
257 changing the dirtymark, so are willing to disturb the downstream hosts.
259 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
261 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
265 https://rt.cpan.org/Ticket/Display.html?id=41199
267 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
269 Strangest output being HAKANARDO who managed to upload
271 Here is a better oneliner that includes also the first line of each
274 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
276 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
278 * memory leak in the syncher? It currently weighs 100M.
282 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
285 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
287 * lookup by epoch and by path and use this ability on the pause to never
288 again register a file twice that doesn't need it. Let's call it
291 * after the dirtymark is done: fill up recentfiles with fake (historic)
292 entries; fill up with individual corrections; algorithm maybe to be done
293 with bigfloat so that we can always place something in the middle
294 between two entries. Before we must switch to bigfloat we could try to
295 use Data::Float::nextup to get the.
297 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
300 * dirtymark now lives in Recentfile, needs to be used in rmirror.
302 * find out why the downloader died after a couple of hours without a net
303 connection. Write a test that survives the not-existence of the other
306 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
308 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
309 it makes less sense with tight coupling for secondary files. But asking
310 the server what the current epoch is might be cheaper on HTTP than on
311 rsync. (Needs to be evaluated)
313 * remove the 0.00 from the verbose overview in the Merged column in the
316 * write tests that expose the problems of the last few days: cascading
317 client/server roles, tight coupling for secondary RFs, deletes after
320 * Some day we might want to have policy options for the slave:
321 tight/loose/no coupling with upstream for secondary RFs. tight is what
322 we have now. loose would wait until a gap occurs that can be closed.
324 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
326 * revisit all $rfs->[$i+1] places if they now make sense still
328 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
330 * another bug is the fact that the mirror command deletes files before
331 it unhides the index file, thus confusing downstream slaves. We must not
332 delete before unhiding and must delete after unhiding. FIXED.
334 * new complication about the slave that is playing a server role.
335 Currently we mirror from newest to oldest with a hidden temporary file
336 as index. And when one file is finished, we unhide the index file.
337 Imagine the cascading server/slave is dead for a day. It then starts
338 mirroring again with the freshest thing and unhides the freshest index
339 file when it has worked through it. In that moment it exposes a time
340 hole. Because it now works on the second recentfile which is still
343 We currently do nothing special to converge after such a drop out. At
344 least not intentionally and robustly and thought through.
346 The algorithm we use to seed the next file needs quite a lot of more
347 robustness than it currently has. Something to do with looking at the
348 merged element of the next rf and when it has dropped off, we seed
349 immediately. And if it ramains dropped off, we seed again, of course.
351 Nope, looking from smaller to larger RFS we look at the merged element
352 of this RF and at the minmax/max element of the next RF. If that
353 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
356 And we need a public accessor seed and unseed or seeded. But not the mix
357 of public and private stuff that then is used behind the back.
359 And then the secondary* stuff must go.
361 And we must understand what the impact is on the DONE system. Can it go
362 unnoticed that there was a hole? And could the DONE system have decided
363 the hole is covered? This should be testable with three directories where
364 the middle stops working for a while. Done->merge is suspicious, we must
365 stop it from merging non-conflatable neighbors due to broken continuity.
369 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
371 * Slaven suggests to have the current epoch or the whole current
372 recentfile available from the HTTP server and take it away with
373 keepalive. This direction goes the granularity down to subseconds.
375 We might want to rewrite everything to factor out transport and allow
376 the whole thing to run via HTTP.
378 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
380 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
383 * smoker on k81 fetching from k75 to verify cascading works. See
384 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
386 * maybe the loop should wait for CHECKSUMS file after every upload. And
387 CPAN.pm needs to deal with timestamps in the future.
389 * do not forget the dirtymark!
391 Text: have a new flag on recentfiles with the meaning: if this
392 changes, you're required to run a full rsync over all the files. The
393 reason why we set it would probably be: some foul happened. we injected
394 files in arbitrary places or didn't inject them although they changed.
395 The content of the flag? Timestamp? The relation between the
396 recentfiles would have to be inheritance from the principal, because any
397 out of band changes would soon later propagate to the next recentfile.
399 By upping the flag often one can easily ruin the slaves.
401 last out of band change? dirtymark?
403 Anyway, this implies that we read a potentially existing recentfile
406 And it implies that we have an eventloop that keeps us busy in 2-3
407 cycles, one for current stuff (tight loop) and one for the recentfiles
408 (cascade when principal has changed), one for the old stuff after a
411 And it implies that the out-of-band change in any of the recentfiles
412 must have a lock on the principal file and there is the place to set the
415 * start a FAQ, especially quick start guide questions. Also to aid those
416 problematic areas where we have no good solution, like the "links"
419 * wish feedback when we are slow.
423 * Remove a few DEBUG statements.
425 * The multiple-rrr way of doing things needs a new option to rmirror,
426 like piecemeal or so. Not urgent because after the first pass through,
427 things run smoothely. It's only ugly during the first pass.
429 * I have the suspicion that the code is broken that decides if the
430 neighboring RF needs to be seeded. I fear when too much time has gone
431 between two calls (in our case more than one hour), it would not seed
432 the neighbor. Of course this will never be noticed, so we need a good
435 * local/localroot confusion: I currently pass both options but one must
438 * accounts for early birds on PAUSE rsync daemon.
440 * hardcoded 20 seconds
442 * who mirrors the index? DOING now.
444 * which CPAN mirrors offer rsync?
446 * visit all XXX, visit all _float places
448 * rename the pathdb stuff, it's too confusing. No idea how.
450 * rrr-inotify, backpan, rrr-register
452 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
454 * current bugs: the pathdb seems to get no reset, the seeding of the
455 secondaryttl stuff seems not to have an effect. Have helped myself with
456 a rand(10), need to fix this back. So not checked in. Does the rand
459 The rand thing helps. The secondaryttl stuff was in the wrong line,
462 The pathdb stuff was because I called either _pathdb or __pathdb on the
463 wrong object. FIXED now.
465 * It's not so beautiful if we never fetch the recentfiles that are not
466 the principal, even if this is correct behaviour. We really do not need
467 them after we have fetched the whole content.
469 OK, we want a switch for that: secondaryttl DONE
471 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
473 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
474 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
475 it lets 40 follow. FIXED
477 * See that the long running process really only updates the principal
478 file unless it has missed a timespan during which something happened. If
479 nothing happened, it must notice even when it misses the timespan. DONE
481 * we must throw away the pathdb when we have reached the end of Z. From
482 that moment we can have a very small pathdb because the only reason for
483 a pathdb is that we know to ignore old records in old files. We won't
484 need this pathdb again before the next full pass over the data is
485 necessary and then we will rebuild it as we go along. DONE
487 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
489 * I think, Done::register_one is doing wrong in that it does not
490 conflate neighboring pieces. The covered() method cannot do this because
491 it has no recent_events array at hand. But register_one has it and could
492 do it and for some reason misses to do it (sometimes).
494 This means that the three tests I just wrote can probably not survive
495 because they test with an already broken Done structure.
497 The art now is to detect how it happens, then to reproduce, then write a
500 So from the logfile this is what happens: we have a good interval with
501 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
502 top of it. Locally we now mirror F2 and open a new done interval for it.
503 Then we mirror F1 but this time with the timestamp T1b. And when we then
504 try to close the gap, we do not find T1 but instead something older. We
505 should gladly accept this older piece and this would fix this bug.
509 * bug to fix: when the 1h file changes while rmirror is running, we do
510 correctly sync the new files but never switch to the 6h file but rather
511 stay in a rather quick loop that fetches the 1h file again and again.
513 Is it possible that we initialize a new object? Or does
514 get_remote_recentfile_as_tempfile overwrite something in myself?
516 Want a new option: _runstatusfile => $file which frequently dumps the
517 state of all recentfiles to a file.
521 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
523 * Todo: now teach update to verify the timestamp is about to write
524 against the previous and use _increase_a_bit if it doesn't comply with
525 strict monotony. DONE
527 * The problem of rounding. So far perl's default precision was
528 sufficient. One day it won't be. FakeFloat has an easy job when it is
529 only reading and other machines have written correctly. But when we want
530 to write a floating point number that is a bit larger than the other
531 one, then we need our own idea of precision.
533 Slaven said: just append a "1". This might be going towards the end of
534 usability too quickly. I'd like something that actually uses the decimal
535 system. Well, appending a 1 also does this but...
537 E.g. we have 1.0. nextup on this architecture is starting with
538 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
539 taken the 1.0000000000000003 and the next user comes and the time tells
540 him 1.0 again. He has to beat my number without stepping over the
541 nextup. This is much less space than I had when I chose 1,2,3.
543 What is also irritating is that nextup is architecture dependent. The
544 128 bit guy must choose very long numbers to fit in between whereas the
545 other one with 16 bit uses larger steps. But then the algorithm is the
546 same for both, so that would be a nice thing.
548 I see two situation where we need this. One is when Time::HiRes returns
549 us a value that is <= the last entry in our recentfile. In this case
550 (let's call it the end-case) we must fill the region between that number
551 and the next higher native floating point number. The other is when we
552 inject an old file into an old recentfile (we would then also set a new
553 dirtymark). We find the integer value already taken and need a slightly
554 different one (let's call it the middle-case). The difference between
555 the two situations is that the next user will want to find something
556 higher than my number in the end-case and something lower than my number
559 So I suggest we give the function both a value and an upper bound and it
560 calculates us a primitive middle. The upper bound in the middle-case is
561 the next integer. The upper bound on the end-case is the nextup floating
562 point number. But the latter poses another problem: if we have occupied
563 the middle m between x and nextup(x), then the nextup(m) will probably
564 not be the same as nextup(x) because some rounding will take place
565 before the nextup is calculated and when the rounding reaches the
566 nextup(x), we will end up at nextup(nextup(x)).
568 So we really need to consider the nextup and the nextdown from there and
569 then the middle and that's the number we may approach asymptotically.
572 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
574 * consider deprecating the use of RECENT.recent as a symlink. It turns
575 out to need extra hoops with the rsync options and just isn't worth it.
576 Or maybe these extra hoops are needed anyway for the rest of the tree?
577 Nope, can't be the case because not all filesystems support symlinks.
579 But before doing the large step, I'll deprecate the call of
580 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
581 only introduced to resolve RECENT.recent and complicates the routine far
582 beyond what it deserves.
584 DONE. Won't deprecate RECENT.recent, just moved its handling to the
587 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
589 * I think it's a bug that the rsync_option links must be set to true in
590 order to support RECENT.recent and that nobody cares to set it
591 automatically. Similar for ignore_link_stat_errors.
593 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
595 * Todo: find all todos together and make a plan what is missing for a
598 - verifytree or something like that. fsck maybe.
600 - rersyncrecent, the script itself? What it do?
602 - a way to only mirror the recentfiles without mirroring the whole
603 remote system such that people can decide to mirror only partially see
604 also 2008-08-30. .shadow-xxx directory? this also needed for a
605 filesystem that is still incomplete and might need the mirrorfiles for
608 - long living objects that mirror again and again. Inject something
609 into ta, see how it goes over to tb.
611 - how do we continue filling up the DONE system when we use an object
612 for the second time? "fully covered" and "uptodate" or new terminology.
614 - overview called on the wrong file should be understandable
616 - the meta data field that must change when we fake something up so that
617 the downstream people know they have to re-fetch everything.
619 - how tolerant are we against missing files upstream? how do we keep
620 track? there are legitimate cases where we did read upstream index right
621 before a file got deleted there and then find that file as new and want
622 it. There are other cases that are not self healing and must be tracked
625 - how, exactly, do we have to deal with deletes? With rsync errors?
627 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
628 authors) failed: No such file or directory (2)
630 The file above is a delete in 1h and a new in file 1M and the
631 delete in the locally running rmirror did not get propagated to the 1M
632 object. Bug. And the consequence is a standstill.
634 It seems that a slave that works with a file below the principal needs
635 to merge things all the way up to get rid of later deletes. Or keep
636 track of all deletes and skip them later. So we need a trackdeletes.pm
637 similar to the done.pm?
639 see also 2008-08-20 about spurious deletes that really have no add
640 counterpart and yet they are not wrong.
642 - consider the effect when resyncing the recentfile takes longer than
643 the time per loop. Then we never rsync any file. We need to diagnose
644 that and force an increase of that loop time. But when we later are fast
645 enough again because the net has recovered, then we need to switch back
646 to original parameters. ERm, no, it's enough to keep syncing at least
647 one file before refetching an index file.
649 - remember to verify that no temp files are left lying around and the
652 - status file for not long running jobs that want to track upstream with
655 - revisit all XXX _float areas and study Sub::Exporter DONE
657 - persistent DB even though we just said we do not need it. Just for
658 extended capabilities and time savings when, for example, upstream
659 announces a reset and we get new recentfiles and could then limit
660 ourselves to a subset of files (those that have a changed epoch) in a
661 first pass and would only then do the loop to verify the rest. Or
664 * Todo: aggregate files should know their feed and finding the principal
665 should be done stepwise. (?)
667 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
668 left locks around. DONE
670 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
672 * maybe extend the _overview so that it always says if and where the
673 last file is in the next file and where the next event in the next rf
674 would lie. No, don't like this anymore. REJECT
676 * take the two new redundant tests out again, only the third must
679 * Todo: add a sanity check if the merged structure is really pointing to
680 a different rf and that this different rf is larger. DONE
682 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
684 * now test, if they are overlapping. And test if there is a file in the
685 next rf that would fit into this rf's interval.
687 1h 1222324012.8474 1222322541.7963 0.4086
688 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
689 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
690 1W 1222313218.3626 1221708477.5829 167.9835
692 I suspect that somebody writes a merged timestamp without having merged
693 and then somebody else relies on it.
695 If aggregate is running, the intervals must not be extravagated, if it
696 is not running, there must not be bounds, the total number of events in
697 the system must be counted and must be controlled throughout the tests.
698 That the test required the additional update was probably nonsense,
699 because aggregate can cut pieces too. FIXED & DONE
701 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
703 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
706 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
708 * Most apparent bug at the moment is that the recentfiles are fetched
709 too often. Only the principal should be fetched and if it has not
710 changed, the others should not be refetched. ATM I must admit that I'm
711 happy that we refetch more often than needed because I can more easily
712 fix bugs while the thing is running.
714 * Let's say, 1220474966.19501 is a timestamp of a file that is already
715 done but the done system does not know about it. The reason for the
716 failure is not known and we never reach the status uptodate because of
717 this. We must get over it.
719 Later it turns out that the origin server had a bug somewhere.
720 1220474966.19042 came after 1220474966.19501. Or better: it was in the
721 array of the recentfile one position above. The bug was my own.
723 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
725 * There is the race condition where the server does a delete and the
726 slave does not yet know and then tries to download it because he sees
727 the new. So for this time window we must be more tolerant against
728 failure. If we cannot download a file, we should just skip it and should
729 not retry immediately. The whole system should discover the lost thing
730 later. Keeping track with the DONE system should really be a no brainer.
732 But there is something more: the whole filesystem is a database and the
733 recentfiles are one possible representation of it. It's a pretty useful
734 representation I think that's why I have implemented something around
735 it. But for strictly local operation it has little value. For local
736 operation we would much rather have a database. So we would enter every
737 recentfile reading and every rsync operation and for every file the last
738 state change and what it leads to. Then we would always ignore older
739 records without the efforts involved with recentfiles.
741 The database would have: path,recentepoch,rsyncedon,deletedon
743 Oh well, not yet clear where this leads to.
745 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
747 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
748 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
750 * statusfile, maybe only on demand, alone to have a sharp debugging
751 tool. It is locked and all recentfiles dump themselves into it and we
752 can build a viewer that lets us know where we stand and what's inside.
754 * remember: only the principal recentfile needs expiration, all others
755 shall be expired by principal if it discovers that something has move
758 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
760 * Always check if we stringify to a higher value than in the entry
763 * And in covered make an additional check if we would be able to see a
764 numerical difference between the two numbers and if we can't then switch
765 to a different, more expensive algorithm. Do not want to be caught by
766 floating surprises. DONE
768 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
770 * caching has several aspects here: we can cache the interval of the
771 recentfile which only will change when the mtime of the file changes. We
772 must re-mirror the recentfile when its ttl has expired. Does have_read
773 tell you anything? It counts nothing at all. Only the mtime is
774 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
775 thing the have_mirrored because it is unrelated to the mtime.
777 * Robustness of floating point calculations! I always thought that the
778 string calculated by the origin server for the floating representation
779 of the epoch time is just a string. When we convert it to a number and
780 later back to a string, the other computer might come to a different
781 conclusion. This must not happen, we want to preserve it under any
782 circumstances. I will have to write tests with overlong sequences that
783 get lost in arithmetic and must see if all still works well. DONE
785 But one fragile point remains: if one host considers a>b and the other
786 one considers them == but no eq. To prevent this, we must probably do
787 some extra homework. DONE
789 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
791 * the concept of tracking DONE needs an object per recentfile that has
792 something like these methods:
794 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
796 covered() register() covered()
798 The unclear thing is how we translate points in time into intervals. We
799 could pass a reference to the current recent_events array when running
800 we_have(xxx) and let the DONE object iterate over it such that it only
801 has to store a list of intervals that can melt into each other. Ah, even
802 passing the list together with a list of indexes seems feasiable.
804 Or maybe ask for the inverted list?
806 Whenever the complete array is covered by the interval we say we are
807 fully covered and if the recentfile is not expired, we are uptodate.
809 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
811 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
813 * need a way to "return" the next entry after the end of a list. When
814 the caller says "before" or "after" we would like to know if he could
815 cover that interval/threshold or not because this influences the effect
816 of a newer timestamp of that recentfile. DONE with $opt{info}.
818 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
820 * one of the next things to tackle: the equivalent of csync2 -TIXU.
822 loop implies tixu (?). Nope, something like --statefile decides. Per
825 T test, I init, X including removals, U nodirtymark
827 So we have no concept of dirtymarks, we only trust that since we are
828 running we have observed everything steadily. But people will not let
829 this program run forever so we must consider both startup penalty and
830 book keeping for later runs. We keep this for later. For now we write a
831 long running mirror that merges several intervals.
833 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
835 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
838 * rersyncrecent, the script: default to one week. The name of the switch
839 is --after. Other switches? --loop!
841 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
843 * need a switch --skip-deletes (?)
845 * need a switch --enduser that tells us that the whole tempfile
846 discipline is not needed when there is no downstream user. (?)
848 Without this switch we cannot have a reasonable recent.pl that just
849 displays the recent additions. Either we accept to download everything.
850 Or we download temporary files without the typical rsync protocol
853 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
854 File::Temp::tempdir, this might be a win.
856 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
858 * apropos missing: we have no push, we never know the downstream
859 servers. People who know their downstream hosts and want to ascertain
860 something will want additional methods we have never thought about, like
861 update or delete a certain file.
863 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
865 * tempted to refactor rmirror into resolve_symlink, localize, etc.
866 Curious if rsync_options=links equal 0 vs. 1 will make the expected
869 * rsync options: it's a bit of a pain that we usually need several rsync
870 options, like compress, links, times, checksum and that there is no
871 reasonable default except the original rsync default. I think wee can
872 safely assume that the rsync options are shared between all recentfile
873 instances within one recent tree.
875 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
877 * deletes: if a delete follows an add quickly enough it may happen that
878 a downstream mirror did not see the add at all! It seems this needs to
879 be mentioned somewhere. The point here is that even if the downstream is
880 never missing the principal timeframe it may encounter a "delete" that
881 has no complimentary "add" anywhere.
883 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
885 * I suspect the treat of metadata is incorrect during read or something.
886 The bug that I am watching is that between 06:08 and 06:09 the 6h file
887 contained more than 6 hours worth of data. At 06:08 we merged into the
888 1d file. We need to take snapshots of the 6h file over the course of an
889 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
892 Much worse: watching the 1h file: right at the moment (at 06:35) it
893 covers 1218867584-1219120397 which is 70 hours.
895 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
896 08:19:44 2008, that is when I checked out last time, so it seems to be
897 aggregating and never truncating?
899 No, correct is: it is never truncating; but wrong is: it is aggregating.
900 It does receive a lot of events from time to time from a larger file.
901 Somehow a large file gets merged into the small one and because the
902 "meta/merged" attribute is missing, nobody is paying attention. I
903 believe that I can fix this by making sure that metadata are honoured
904 during read. DONE and test adjusted.
906 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
908 * grand renaming plan
910 remotebase => remoteroot to fit well with localroot DONE
911 local_path() => localroot seems to me should already work DONE
912 recentfile_basename => rfilename no need to stress it has no slash DONE
914 filenameroot??? Doesn't seem too bad to me today. Maybe something like
915 kern? It would anyway need a deprecation cycle because it is an
916 important constructor.
918 * I like the portability that Data::Serializer brings us but the price
919 is that some day we might find out that it is slowing us a bit. We'll
922 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
924 * should we not enter the interval of the principal (or the interval of
925 the merging file?) in every aggregated/merged file?
927 * we should aim at a first release and give up on thinking about
928 sanitizing stuff and zloop. Let's just admit that a full traditional
929 rsync is the only available sanitizer ATM. Otherwise it's complicated
930 stuff: sanitizing on the origin server, sanitizing on the slaves,
931 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
932 get the basics out before this becomes a major cause for mess.
934 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
936 * On OSes not supporting symlinks we expect that RECENT.recent contains
937 the contents of the principal recentfile. Actually this is identical on
938 systems supporting symlinks. Simple, what follows from that is that we
939 need to keep the serializer in the metadata because we cannot read it
940 from the filename, doesn't it? Of course not. It's a chicken and egg
941 problem. This leaves us with the problem to actually parse the
942 serialized data to find out in which format it is. So who can do the 4
943 or 5 magics we wanted to support? File::LibMagic?
945 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
947 * remotebase and recentfile_basename are ugly names. Now that we need a
948 word for the shortest/principal/driving recentfile too we should do
951 localroot is good. rfile is good. local_path() is bad, local_path($path)
952 is medium, filenameroot() is bad, remotebase is bad, recentfile is
955 Up to now remotebase was the string that described the remote root
956 directory in rsync notation, like pause.perl.org::authors. And
957 recentfile_basename was "RECENT-1h.yaml".
959 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
961 * The test that was added in today's checkin is a good start for a test
962 of rmirror. We should have more methods in Recent.pm: verify,
963 addmissingfiles. We should verify the current tree, then rmirror it and
964 then verifytree the copy. We could then add some arbitrary file and let
965 it be discovered by addmissingfiles, then rmirror again and then
966 verifytree the copy again.
968 Then we could start stealing from csync2 sqlite database [no port to
969 OSX!] and fill a local DB. And methods to compare the database with the
970 recentfiles. Our strength is that in principle we could maintain state
971 with a single float. We have synced up to 1234567890.123456. If the Z
972 file does not add new files all we have to do is mirror the new ones and
975 This makes it clear that we should extend current protocol and declare
976 that we cheat when we add files too late, just to help the other end
977 keeping track. Ah yes, that's what was meant when zloop was mentioned
980 Maybe need to revisit File::Mirror to help me with this task.
982 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
984 * There must be an allow-me-to-truncate flag in every recentfile.
985 Without it one could construct a sequence of updates winning the locking
986 battle against the aggregator. Only if an aggregator has managed to
987 merge data over to the next level, truncating can be allowed. DONE with
990 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
992 * We should probably guarantee that no duplicates enter the aggregator
995 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
997 * To get merge operation faster would need a good benchmark test. What
998 02 spits out isn't reliable enough and is dominated by many other
1001 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1002 Author: Andreas J. Koenig <k@k75.(none)>
1003 Date: Sat Aug 2 07:58:04 2008 +0200
1007 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1008 Author: Andreas J. Koenig <k@k75.(none)>
1009 Date: Sat Aug 2 11:40:29 2008 +0200
1011 there was a lot of trying but the effect is hardly measurable with
1014 * overhead of connecting seems high. When setting
1015 max_files_per_connection to 1 we see that.
1017 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1019 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1021 25138.2215600014/3600 = 6.98283932222261
1023 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1024 be a bug in the aggregator. FIXED
1026 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1028 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1029 we have it, can we mirror it, mirror it!
1031 I fear this needs a new class which might be called
1032 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1033 and treat them as an entity. I realize that a single recentfile may be
1034 sufficient for certain tasks and that it is handy for the low level
1035 programmer but it is not nice to use. If there is a delete in the 1h
1036 file then the 6h file still contains it. Seekers of the best information
1037 need to combine at least some of the recentfiles most of the time.
1039 There is the place for the Z loop!
1041 But the combination is something to collect in a database, isn't it. Did
1042 csync2 just harrumph?
1044 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1046 * it just occurred to me that hosts in the same mirroring pool could
1047 help out each other even without rewriting the recentfile. Just fetch
1048 the stuff to mirror from several places, bingo. But that's something
1049 that should rather live in a separate package or in rsync directly.
1051 * cronjobs are unsuited because with ntp they would all come at the full
1052 minute and disturb each other. Besides that I'd hate to have a backbone
1053 with more than a few seconds latency.
1055 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1057 * a second rsync server with access control for PAUSE. Port? 873 is the
1058 standard port, let's take 8873.
1060 * if there were a filesystem based on this, it would have a slow access
1061 to inexistent files. It would probably provide wrong readdir (only based
1062 on current content) or also a slow one (based on a recentfile written
1063 after the call). But it would provide fast access to existing files. Or
1064 one would deliberately allow slightly blurred answers based on some
1065 sqlite reflection of the recentfiles.
1067 * todo: write a variant of mirror() that combines two or more
1068 recentfiles and treats them like one
1070 * todo: signal handler to remove the tempfile
1072 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1074 * now that we have the symlink I forgot how it should be used in
1077 * the z loop: add missing files to Z file. Just append them (instead of
1078 prepending). So one guy prepends something from the Y file from time to
1079 time and another guy appends something rather frequently. Collecting
1080 pond. When Y merges into Z, things get epoch and the collecting pond
1081 gets smaller. What exactly are "missing files"?
1083 take note of current epoch of the alpha file, let's call it the
1086 find all files on disk
1088 remove all files registered in the recentworld up to recent-ts
1090 remove all files that have been deleted after recent-ts according to
1093 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1095 * rersyncrecent might be a cronjob with a (locked) state file which
1096 contains things like after and maybe last z sync or such?
1098 rrr-mirror might be an alternative name but how would we justify the
1099 three Rs when there is no Re-Rsync-Recent?
1101 With the --loop parameter it is an endless loop, without it is no loop.
1102 At least this is simple.
1104 * todo: new accssor z-interval specifies how often the Z file is updated
1105 against the filesystem. We probably want no epoch stamp on these
1106 entries. And we want to be able to filter the entries (e.g. no
1107 by-modules and by-category tree)
1109 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1111 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1112 into Z. We do not need the other files filled up with filesystem
1115 * need interface to query for a file in order to NOT call update on
1116 PAUSE a second time within a short time.
1118 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1120 * recommended update interval? Makes no sense, is different for
1127 change-log-default-name: "Todo"