1 2009-03-17 Andreas J. Koenig <andk@cpan.org>
3 * done: verified the existence of the floating point bug in bleadperl
5 * todo: integrate a dirty update with two aggregate calls before
6 unlocking for frictionless dirtying
8 * todo: start the second rsync daemon on pause
10 * todo: move index files to .recent
12 2009-03-16 Andreas J. Koenig <andk@cpan.org>
14 * What was the resolution of the mirror.pl delete hook bug? Do we call
15 the delete hook when pause removes a file from MUIR?
17 * Today on pause: Updating 2a13fba..29f284d and installing it for
18 /usr/local/perl-5.10.0{,-RC2}
20 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
21 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
23 2009-03-15 Andreas J. Koenig <andk@cpan.org>
25 * currently recent_events has the side effect of setting dirtymark
26 because it forces a read on the file. That should be transparent, so
27 that the dirtymark call always forces a cache-able(?) read.
29 * The bug below is -- after a lot of trying -- not reproducible on a
30 small script, only in the large test script. The closest to the output
36 my $x = "01237123229.8814";
41 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
43 $l = "1237123231.22458";
44 $r = "1237123231.22458";
48 Devel::Peek::Dump $x = $l <=> $r;
52 The checked in state at c404a85 fails the test with my
53 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
54 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
55 in the test script always consistent.
57 * Todo: write a test that inserts a second dirty file with an already
58 existing timestamp. DONE
60 * Bug in perl 5.10 on my 64bit box:
62 DB<98> Devel::Peek::Dump $l
63 SV = PVMG(0x19e0450) at 0x142a550
65 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
68 PV = 0x194ce70 "1237123231.22458"\0
72 DB<99> Devel::Peek::Dump $r
73 SV = PVMG(0x19e0240) at 0x142a3e8
75 FLAGS = (PADMY,POK,pPOK)
78 PV = 0x19ff900 "1237123231.22458"\0
82 DB<100> Devel::Peek::Dump $l <=> $r
83 SV = IV(0x19ea6e8) at 0x19ea6f0
85 FLAGS = (PADTMP,IOK,pIOK)
88 DB<101> Devel::Peek::Dump $l
89 SV = PVMG(0x19e0450) at 0x142a550
91 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
94 PV = 0x194ce70 "1237123231.22458"\0
98 DB<102> Devel::Peek::Dump $r
99 SV = PVMG(0x19e0240) at 0x142a3e8
101 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
103 NV = 1237123231.22458
104 PV = 0x19ff900 "1237123231.22458"\0
108 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
110 * Todo: reset "done" or "covered" and "minmax" after a dirty operation?
112 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
114 * $obj->merge ($other) needs to learn about equal epoch which may happen
115 since dirty_epoch intruded.
117 * Wontfix anytime soon: I think we currently do not support mkdir. Only
120 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
122 * Todo: continue working on update(...,$dirty_epoch). It must be
123 followed by a fast_aggregate!
125 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
127 * maybe we need a closest_entry or fitting_interval or something like
128 that. We want to merge an event into the middle of some recentfile.
129 First we do not know which file, then we do not know where to lock,
130 where to enter the new item, when and where to correct the dirtymark.
132 So my thought is we should first find which file.
134 Another part of my brain answers: what would happen if we would enter
135 the new file into the smallest file just like an ordinary new event,
136 just as an old event?
138 (1) we would write a duplicate timestamp? No, this would be easy to
141 (2) we would make the file large quickly? Yes, but so what? We are
142 changing the dirtymark, so are willing to disturb the downstream hosts.
144 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
146 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
150 https://rt.cpan.org/Ticket/Display.html?id=41199
152 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
154 Strangest output being HAKANARDO who managed to upload
156 Here is a better oneliner that includes also the first line of each
159 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
161 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
163 * memory leak in the syncher? It currently weighs 100M.
167 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
170 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
172 * lookup by epoch and by path and use this ability on the pause to never
173 again register a file twice that doesn't need it. Let's call it
176 * after the dirtymark is done: fill up recentfiles with fake (historic)
177 entries; fill up with individual corrections; algorithm maybe to be done
178 with bigfloat so that we can always place something in the middle
179 between two entries. Before we must switch to bigfloat we could try to
180 use Data::Float::nextup to get the.
182 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
185 * dirtymark now lives in Recentfile, needs to be used in rmirror.
187 * find out why the downloader died after a couple of hours without a net
188 connection. Write a test that survives the not-existence of the other
191 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
193 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
194 it makes less sense with tight coupling for secondary files. But asking
195 the server what the current epoch is might be cheaper on HTTP than on
196 rsync. (Needs to be evaluated)
198 * remove the 0.00 from the verbose overview in the Merged column in the
201 * write tests that expose the problems of the last few days: cascading
202 client/server roles, tight coupling for secondary RFs, deletes after
205 * Some day we might want to have policy options for the slave:
206 tight/loose/no coupling with upstream for secondary RFs. tight is what
207 we have now. loose would wait until a gap occurs that can be closed.
209 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
211 * revisit all $rfs->[$i+1] places if they now make sense still
213 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
215 * another bug is the fact that the mirror command deletes files before
216 it unhides the index file, thus confusing downstream slaves. We must not
217 delete before unhiding and must delete after unhiding. FIXED.
219 * new complication about the slave that is playing a server role.
220 Currently we mirror from newest to oldest with a hidden temporary file
221 as index. And when one file is finished, we unhide the index file.
222 Imagine the cascading server/slave is dead for a day. It then starts
223 mirroring again with the freshest thing and unhides the freshest index
224 file when it has worked through it. In that moment it exposes a time
225 hole. Because it now works on the second recentfile which is still
228 We currently do nothing special to converge after such a drop out. At
229 least not intentionally and robustly and thought through.
231 The algorithm we use to seed the next file needs quite a lot of more
232 robustness than it currently has. Something to do with looking at the
233 merged element of the next rf and when it has dropped off, we seed
234 immediately. And if it ramains dropped off, we seed again, of course.
236 Nope, looking from smaller to larger RFS we look at the merged element
237 of this RF and at the minmax/max element of the next RF. If that
238 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
241 And we need a public accessor seed and unseed or seeded. But not the mix
242 of public and private stuff that then is used behind the back.
244 And then the secondary* stuff must go.
246 And we must understand what the impact is on the DONE system. Can it go
247 unnoticed that there was a hole? And could the DONE system have decided
248 the hole is covered? This should be testable with three directories where
249 the middle stops working for a while. Done->merge is suspicious, we must
250 stop it from merging non-conflatable neighbors due to broken continuity.
254 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
256 * Slaven suggests to have the current epoch or the whole current
257 recentfile available from the HTTP server and take it away with
258 keepalive. This direction goes the granularity down to subseconds.
260 We might want to rewrite everything to factor out transport and allow
261 the whole thing to run via HTTP.
263 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
265 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
268 * smoker on k81 fetching from k75 to verify cascading works. See
269 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
271 * maybe the loop should wait for CHECKSUMS file after every upload. And
272 CPAN.pm needs to deal with timestamps in the future.
274 * do not forget the dirtymark!
276 Text: have a new flag on recentfiles with the meaning: if this
277 changes, you're required to run a full rsync over all the files. The
278 reason why we set it would probably be: some foul happened. we injected
279 files in arbitrary places or didn't inject them although they changed.
280 The content of the flag? Timestamp? The relation between the
281 recentfiles would have to be inheritance from the principal, because any
282 out of band changes would soon later propagate to the next recentfile.
284 By upping the flag often one can easily ruin the slaves.
286 last out of band change? dirtymark?
288 Anyway, this implies that we read a potentially existing recentfile
291 And it implies that we have an eventloop that keeps us busy in 2-3
292 cycles, one for current stuff (tight loop) and one for the recentfiles
293 (cascade when principal has changed), one for the old stuff after a
296 And it implies that the out-of-band change in any of the recentfiles
297 must have a lock on the principal file and there is the place to set the
300 * start a FAQ, especially quick start guide questions. Also to aid those
301 problematic areas where we have no good solution, like the "links"
304 * wish feedback when we are slow.
308 * Remove a few DEBUG statements.
310 * The multiple-rrr way of doing things needs a new option to rmirror,
311 like piecemeal or so. Not urgent because after the first pass through,
312 things run smoothely. It's only ugly during the first pass.
314 * I have the suspicion that the code is broken that decides if the
315 neighboring RF needs to be seeded. I fear when too much time has gone
316 between two calls (in our case more than one hour), it would not seed
317 the neighbor. Of course this will never be noticed, so we need a good
320 * local/localroot confusion: I currently pass both options but one must
323 * accounts for early birds on PAUSE rsync daemon.
325 * hardcoded 20 seconds
327 * who mirrors the index? DOING now.
329 * which CPAN mirrors offer rsync?
331 * visit all XXX, visit all _float places
333 * rename the pathdb stuff, it's too confusing. No idea how.
335 * rrr-inotify, backpan, rrr-register
337 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
339 * current bugs: the pathdb seems to get no reset, the seeding of the
340 secondaryttl stuff seems not to have an effect. Have helped myself with
341 a rand(10), need to fix this back. So not checked in. Does the rand
344 The rand thing helps. The secondaryttl stuff was in the wrong line,
347 The pathdb stuff was because I called either _pathdb or __pathdb on the
348 wrong object. FIXED now.
350 * It's not so beautiful if we never fetch the recentfiles that are not
351 the principal, even if this is correct behaviour. We really do not need
352 them after we have fetched the whole content.
354 OK, we want a switch for that: secondaryttl DONE
356 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
358 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
359 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
360 it lets 40 follow. FIXED
362 * See that the long running process really only updates the principal
363 file unless it has missed a timespan during which something happened. If
364 nothing happened, it must notice even when it misses the timespan. DONE
366 * we must throw away the pathdb when we have reached the end of Z. From
367 that moment we can have a very small pathdb because the only reason for
368 a pathdb is that we know to ignore old records in old files. We won't
369 need this pathdb again before the next full pass over the data is
370 necessary and then we will rebuild it as we go along. DONE
372 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
374 * I think, Done::register_one is doing wrong in that it does not
375 conflate neighboring pieces. The covered() method cannot do this because
376 it has no recent_events array at hand. But register_one has it and could
377 do it and for some reason misses to do it (sometimes).
379 This means that the three tests I just wrote can probably not survive
380 because they test with an already broken Done structure.
382 The art now is to detect how it happens, then to reproduce, then write a
385 So from the logfile this is what happens: we have a good interval with
386 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
387 top of it. Locally we now mirror F2 and open a new done interval for it.
388 Then we mirror F1 but this time with the timestamp T1b. And when we then
389 try to close the gap, we do not find T1 but instead something older. We
390 should gladly accept this older piece and this would fix this bug.
394 * bug to fix: when the 1h file changes while rmirror is running, we do
395 correctly sync the new files but never switch to the 6h file but rather
396 stay in a rather quick loop that fetches the 1h file again and again.
398 Is it possible that we initialize a new object? Or does
399 get_remote_recentfile_as_tempfile overwrite something in myself?
401 Want a new option: _runstatusfile => $file which frequently dumps the
402 state of all recentfiles to a file.
406 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
408 * Todo: now teach update to verify the timestamp is about to write
409 against the previous and use _increase_a_bit if it doesn't comply with
410 strict monotony. DONE
412 * The problem of rounding. So far perl's default precision was
413 sufficient. One day it won't be. FakeFloat has an easy job when it is
414 only reading and other machines have written correctly. But when we want
415 to write a floating point number that is a bit larger than the other
416 one, then we need our own idea of precision.
418 Slaven said: just append a "1". This might be going towards the end of
419 usability too quickly. I'd like something that actually uses the decimal
420 system. Well, appending a 1 also does this but...
422 E.g. we have 1.0. nextup on this architecture is starting with
423 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
424 taken the 1.0000000000000003 and the next user comes and the time tells
425 him 1.0 again. He has to beat my number without stepping over the
426 nextup. This is much less space than I had when I chose 1,2,3.
428 What is also irritating is that nextup is architecture dependent. The
429 128 bit guy must choose very long numbers to fit in between whereas the
430 other one with 16 bit uses larger steps. But then the algorithm is the
431 same for both, so that would be a nice thing.
433 I see two situation where we need this. One is when Time::HiRes returns
434 us a value that is <= the last entry in our recentfile. In this case
435 (let's call it the end-case) we must fill the region between that number
436 and the next higher native floating point number. The other is when we
437 inject an old file into an old recentfile (we would then also set a new
438 dirtymark). We find the integer value already taken and need a slightly
439 different one (let's call it the middle-case). The difference between
440 the two situations is that the next user will want to find something
441 higher than my number in the end-case and something lower than my number
444 So I suggest we give the function both a value and an upper bound and it
445 calculates us a primitive middle. The upper bound in the middle-case is
446 the next integer. The upper bound on the end-case is the nextup floating
447 point number. But the latter poses another problem: if we have occupied
448 the middle m between x and nextup(x), then the nextup(m) will probably
449 not be the same as nextup(x) because some rounding will take place
450 before the nextup is calculated and when the rounding reaches the
451 nextup(x), we will end up at nextup(nextup(x)).
453 So we really need to consider the nextup and the nextdown from there and
454 then the middle and that's the number we may approach asymptotically.
457 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
459 * consider deprecating the use of RECENT.recent as a symlink. It turns
460 out to need extra hoops with the rsync options and just isn't worth it.
461 Or maybe these extra hoops are needed anyway for the rest of the tree?
462 Nope, can't be the case because not all filesystems support symlinks.
464 But before doing the large step, I'll deprecate the call of
465 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
466 only introduced to resolve RECENT.recent and complicates the routine far
467 beyond what it deserves.
469 DONE. Won't deprecate RECENT.recent, just moved its handling to the
472 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
474 * I think it's a bug that the rsync_option links must be set to true in
475 order to support RECENT.recent and that nobody cares to set it
476 automatically. Similar for ignore_link_stat_errors.
478 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
480 * Todo: find all todos together and make a plan what is missing for a
483 - verifytree or something like that. fsck maybe.
485 - rersyncrecent, the script itself? What it do?
487 - a way to only mirror the recentfiles without mirroring the whole
488 remote system such that people can decide to mirror only partially see
489 also 2008-08-30. .shadow-xxx directory? this also needed for a
490 filesystem that is still incomplete and might need the mirrorfiles for
493 - long living objects that mirror again and again. Inject something
494 into ta, see how it goes over to tb.
496 - how do we continue filling up the DONE system when we use an object
497 for the second time? "fully covered" and "uptodate" or new terminology.
499 - overview called on the wrong file should be understandable
501 - the meta data field that must change when we fake something up so that
502 the downstream people know they have to re-fetch everything.
504 - how tolerant are we against missing files upstream? how do we keep
505 track? there are legitimate cases where we did read upstream index right
506 before a file got deleted there and then find that file as new and want
507 it. There are other cases that are not self healing and must be tracked
510 - how, exactly, do we have to deal with deletes? With rsync errors?
512 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
513 authors) failed: No such file or directory (2)
515 The file above is a delete in 1h and a new in file 1M and the
516 delete in the locally running rmirror did not get propagated to the 1M
517 object. Bug. And the consequence is a standstill.
519 It seems that a slave that works with a file below the principal needs
520 to merge things all the way up to get rid of later deletes. Or keep
521 track of all deletes and skip them later. So we need a trackdeletes.pm
522 similar to the done.pm?
524 see also 2008-08-20 about spurious deletes that really have no add
525 counterpart and yet they are not wrong.
527 - consider the effect when resyncing the recentfile takes longer than
528 the time per loop. Then we never rsync any file. We need to diagnose
529 that and force an increase of that loop time. But when we later are fast
530 enough again because the net has recovered, then we need to switch back
531 to original parameters. ERm, no, it's enough to keep syncing at least
532 one file before refetching an index file.
534 - remember to verify that no temp files are left lying around and the
537 - status file for not long running jobs that want to track upstream with
540 - revisit all XXX _float areas and study Sub::Exporter DONE
542 - persistent DB even though we just said we do not need it. Just for
543 extended capabilities and time savings when, for example, upstream
544 announces a reset and we get new recentfiles and could then limit
545 ourselves to a subset of files (those that have a changed epoch) in a
546 first pass and would only then do the loop to verify the rest. Or
549 * Todo: aggregate files should know their feed and finding the principal
550 should be done stepwise. (?)
552 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
553 left locks around. DONE
555 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
557 * maybe extend the _overview so that it always says if and where the
558 last file is in the next file and where the next event in the next rf
559 would lie. No, don't like this anymore. REJECT
561 * take the two new redundant tests out again, only the third must
564 * Todo: add a sanity check if the merged structure is really pointing to
565 a different rf and that this different rf is larger. DONE
567 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
569 * now test, if they are overlapping. And test if there is a file in the
570 next rf that would fit into this rf's interval.
572 1h 1222324012.8474 1222322541.7963 0.4086
573 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
574 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
575 1W 1222313218.3626 1221708477.5829 167.9835
577 I suspect that somebody writes a merged timestamp without having merged
578 and then somebody else relies on it.
580 If aggregate is running, the intervals must not be extravagated, if it
581 is not running, there must not be bounds, the total number of events in
582 the system must be counted and must be controlled throughout the tests.
583 That the test required the additional update was probably nonsense,
584 because aggregate can cut pieces too. FIXED & DONE
586 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
588 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
591 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
593 * Most apparent bug at the moment is that the recentfiles are fetched
594 too often. Only the principal should be fetched and if it has not
595 changed, the others should not be refetched. ATM I must admit that I'm
596 happy that we refetch more often than needed because I can more easily
597 fix bugs while the thing is running.
599 * Let's say, 1220474966.19501 is a timestamp of a file that is already
600 done but the done system does not know about it. The reason for the
601 failure is not known and we never reach the status uptodate because of
602 this. We must get over it.
604 Later it turns out that the origin server had a bug somewhere.
605 1220474966.19042 came after 1220474966.19501. Or better: it was in the
606 array of the recentfile one position above. The bug was my own.
608 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
610 * There is the race condition where the server does a delete and the
611 slave does not yet know and then tries to download it because he sees
612 the new. So for this time window we must be more tolerant against
613 failure. If we cannot download a file, we should just skip it and should
614 not retry immediately. The whole system should discover the lost thing
615 later. Keeping track with the DONE system should really be a no brainer.
617 But there is something more: the whole filesystem is a database and the
618 recentfiles are one possible representation of it. It's a pretty useful
619 representation I think that's why I have implemented something around
620 it. But for strictly local operation it has little value. For local
621 operation we would much rather have a database. So we would enter every
622 recentfile reading and every rsync operation and for every file the last
623 state change and what it leads to. Then we would always ignore older
624 records without the efforts involved with recentfiles.
626 The database would have: path,recentepoch,rsyncedon,deletedon
628 Oh well, not yet clear where this leads to.
630 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
632 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
633 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
635 * statusfile, maybe only on demand, alone to have a sharp debugging
636 tool. It is locked and all recentfiles dump themselves into it and we
637 can build a viewer that lets us know where we stand and what's inside.
639 * remember: only the principal recentfile needs expiration, all others
640 shall be expired by principal if it discovers that something has move
643 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
645 * Always check if we stringify to a higher value than in the entry
648 * And in covered make an additional check if we would be able to see a
649 numerical difference between the two numbers and if we can't then switch
650 to a different, more expensive algorithm. Do not want to be caught by
651 floating surprises. DONE
653 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
655 * caching has several aspects here: we can cache the interval of the
656 recentfile which only will change when the mtime of the file changes. We
657 must re-mirror the recentfile when its ttl has expired. Does have_read
658 tell you anything? It counts nothing at all. Only the mtime is
659 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
660 thing the have_mirrored because it is unrelated to the mtime.
662 * Robustness of floating point calculations! I always thought that the
663 string calculated by the origin server for the floating representation
664 of the epoch time is just a string. When we convert it to a number and
665 later back to a string, the other computer might come to a different
666 conclusion. This must not happen, we want to preserve it under any
667 circumstances. I will have to write tests with overlong sequences that
668 get lost in arithmetic and must see if all still works well. DONE
670 But one fragile point remains: if one host considers a>b and the other
671 one considers them == but no eq. To prevent this, we must probably do
672 some extra homework. DONE
674 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
676 * the concept of tracking DONE needs an object per recentfile that has
677 something like these methods:
679 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
681 covered() register() covered()
683 The unclear thing is how we translate points in time into intervals. We
684 could pass a reference to the current recent_events array when running
685 we_have(xxx) and let the DONE object iterate over it such that it only
686 has to store a list of intervals that can melt into each other. Ah, even
687 passing the list together with a list of indexes seems feasiable.
689 Or maybe ask for the inverted list?
691 Whenever the complete array is covered by the interval we say we are
692 fully covered and if the recentfile is not expired, we are uptodate.
694 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
696 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
698 * need a way to "return" the next entry after the end of a list. When
699 the caller says "before" or "after" we would like to know if he could
700 cover that interval/threshold or not because this influences the effect
701 of a newer timestamp of that recentfile. DONE with $opt{info}.
703 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
705 * one of the next things to tackle: the equivalent of csync2 -TIXU.
707 loop implies tixu (?). Nope, something like --statefile decides. Per
710 T test, I init, X including removals, U nodirtymark
712 So we have no concept of dirtymarks, we only trust that since we are
713 running we have observed everything steadily. But people will not let
714 this program run forever so we must consider both startup penalty and
715 book keeping for later runs. We keep this for later. For now we write a
716 long running mirror that merges several intervals.
718 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
720 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
723 * rersyncrecent, the script: default to one week. The name of the switch
724 is --after. Other switches? --loop!
726 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
728 * need a switch --skip-deletes (?)
730 * need a switch --enduser that tells us that the whole tempfile
731 discipline is not needed when there is no downstream user. (?)
733 Without this switch we cannot have a reasonable recent.pl that just
734 displays the recent additions. Either we accept to download everything.
735 Or we download temporary files without the typical rsync protocol
738 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
739 File::Temp::tempdir, this might be a win.
741 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
743 * apropos missing: we have no push, we never know the downstream
744 servers. People who know their downstream hosts and want to ascertain
745 something will want additional methods we have never thought about, like
746 update or delete a certain file.
748 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
750 * tempted to refactor rmirror into resolve_symlink, localize, etc.
751 Curious if rsync_options=links equal 0 vs. 1 will make the expected
754 * rsync options: it's a bit of a pain that we usually need several rsync
755 options, like compress, links, times, checksum and that there is no
756 reasonable default except the original rsync default. I think wee can
757 safely assume that the rsync options are shared between all recentfile
758 instances within one recent tree.
760 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
762 * deletes: if a delete follows an add quickly enough it may happen that
763 a downstream mirror did not see the add at all! It seems this needs to
764 be mentioned somewhere. The point here is that even if the downstream is
765 never missing the principal timeframe it may encounter a "delete" that
766 has no complimentary "add" anywhere.
768 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
770 * I suspect the treat of metadata is incorrect during read or something.
771 The bug that I am watching is that between 06:08 and 06:09 the 6h file
772 contained more than 6 hours worth of data. At 06:08 we merged into the
773 1d file. We need to take snapshots of the 6h file over the course of an
774 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
777 Much worse: watching the 1h file: right at the moment (at 06:35) it
778 covers 1218867584-1219120397 which is 70 hours.
780 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
781 08:19:44 2008, that is when I checked out last time, so it seems to be
782 aggregating and never truncating?
784 No, correct is: it is never truncating; but wrong is: it is aggregating.
785 It does receive a lot of events from time to time from a larger file.
786 Somehow a large file gets merged into the small one and because the
787 "meta/merged" attribute is missing, nobody is paying attention. I
788 believe that I can fix this by making sure that metadata are honoured
789 during read. DONE and test adjusted.
791 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
793 * grand renaming plan
795 remotebase => remoteroot to fit well with localroot DONE
796 local_path() => localroot seems to me should already work DONE
797 recentfile_basename => rfilename no need to stress it has no slash DONE
799 filenameroot??? Doesn't seem too bad to me today. Maybe something like
800 kern? It would anyway need a deprecation cycle because it is an
801 important constructor.
803 * I like the portability that Data::Serializer brings us but the price
804 is that some day we might find out that it is slowing us a bit. We'll
807 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
809 * should we not enter the interval of the principal (or the interval of
810 the merging file?) in every aggregated/merged file?
812 * we should aim at a first release and give up on thinking about
813 sanitizing stuff and zloop. Let's just admit that a full traditional
814 rsync is the only available sanitizer ATM. Otherwise it's complicated
815 stuff: sanitizing on the origin server, sanitizing on the slaves,
816 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
817 get the basics out before this becomes a major cause for mess.
819 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
821 * On OSes not supporting symlinks we expect that RECENT.recent contains
822 the contents of the principal recentfile. Actually this is identical on
823 systems supporting symlinks. Simple, what follows from that is that we
824 need to keep the serializer in the metadata because we cannot read it
825 from the filename, doesn't it? Of course not. It's a chicken and egg
826 problem. This leaves us with the problem to actually parse the
827 serialized data to find out in which format it is. So who can do the 4
828 or 5 magics we wanted to support? File::LibMagic?
830 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
832 * remotebase and recentfile_basename are ugly names. Now that we need a
833 word for the shortest/principal/driving recentfile too we should do
836 localroot is good. rfile is good. local_path() is bad, local_path($path)
837 is medium, filenameroot() is bad, remotebase is bad, recentfile is
840 Up to now remotebase was the string that described the remote root
841 directory in rsync notation, like pause.perl.org::authors. And
842 recentfile_basename was "RECENT-1h.yaml".
844 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
846 * The test that was added in today's checkin is a good start for a test
847 of rmirror. We should have more methods in Recent.pm: verify,
848 addmissingfiles. We should verify the current tree, then rmirror it and
849 then verifytree the copy. We could then add some arbitrary file and let
850 it be discovered by addmissingfiles, then rmirror again and then
851 verifytree the copy again.
853 Then we could start stealing from csync2 sqlite database [no port to
854 OSX!] and fill a local DB. And methods to compare the database with the
855 recentfiles. Our strength is that in principle we could maintain state
856 with a single float. We have synced up to 1234567890.123456. If the Z
857 file does not add new files all we have to do is mirror the new ones and
860 This makes it clear that we should extend current protocol and declare
861 that we cheat when we add files too late, just to help the other end
862 keeping track. Ah yes, that's what was meant when zloop was mentioned
865 Maybe need to revisit File::Mirror to help me with this task.
867 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
869 * There must be an allow-me-to-truncate flag in every recentfile.
870 Without it one could construct a sequence of updates winning the locking
871 battle against the aggregator. Only if an aggregator has managed to
872 merge data over to the next level, truncating can be allowed. DONE with
875 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
877 * We should probably guarantee that no duplicates enter the aggregator
880 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
882 * To get merge operation faster would need a good benchmark test. What
883 02 spits out isn't reliable enough and is dominated by many other
886 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
887 Author: Andreas J. Koenig <k@k75.(none)>
888 Date: Sat Aug 2 07:58:04 2008 +0200
892 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
893 Author: Andreas J. Koenig <k@k75.(none)>
894 Date: Sat Aug 2 11:40:29 2008 +0200
896 there was a lot of trying but the effect is hardly measurable with
899 * overhead of connecting seems high. When setting
900 max_files_per_connection to 1 we see that.
902 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
904 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
906 25138.2215600014/3600 = 6.98283932222261
908 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
909 be a bug in the aggregator. FIXED
911 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
913 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
914 we have it, can we mirror it, mirror it!
916 I fear this needs a new class which might be called
917 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
918 and treat them as an entity. I realize that a single recentfile may be
919 sufficient for certain tasks and that it is handy for the low level
920 programmer but it is not nice to use. If there is a delete in the 1h
921 file then the 6h file still contains it. Seekers of the best information
922 need to combine at least some of the recentfiles most of the time.
924 There is the place for the Z loop!
926 But the combination is something to collect in a database, isn't it. Did
927 csync2 just harrumph?
929 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
931 * it just occurred to me that hosts in the same mirroring pool could
932 help out each other even without rewriting the recentfile. Just fetch
933 the stuff to mirror from several places, bingo. But that's something
934 that should rather live in a separate package or in rsync directly.
936 * cronjobs are unsuited because with ntp they would all come at the full
937 minute and disturb each other. Besides that I'd hate to have a backbone
938 with more than a few seconds latency.
940 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
942 * a second rsync server with access control for PAUSE. Port? 873 is the
943 standard port, let's take 8873.
945 * if there were a filesystem based on this, it would have a slow access
946 to inexistent files. It would probably provide wrong readdir (only based
947 on current content) or also a slow one (based on a recentfile written
948 after the call). But it would provide fast access to existing files. Or
949 one would deliberately allow slightly blurred answers based on some
950 sqlite reflection of the recentfiles.
952 * todo: write a variant of mirror() that combines two or more
953 recentfiles and treats them like one
955 * todo: signal handler to remove the tempfile
957 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
959 * now that we have the symlink I forgot how it should be used in
962 * the z loop: add missing files to Z file. Just append them (instead of
963 prepending). So one guy prepends something from the Y file from time to
964 time and another guy appends something rather frequently. Collecting
965 pond. When Y merges into Z, things get epoch and the collecting pond
966 gets smaller. What exactly are "missing files"?
968 take note of current epoch of the alpha file, let's call it the
971 find all files on disk
973 remove all files registered in the recentworld up to recent-ts
975 remove all files that have been deleted after recent-ts according to
978 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
980 * rersyncrecent might be a cronjob with a (locked) state file which
981 contains things like after and maybe last z sync or such?
983 rrr-mirror might be an alternative name but how would we justify the
984 three Rs when there is no Re-Rsync-Recent?
986 With the --loop parameter it is an endless loop, without it is no loop.
987 At least this is simple.
989 * todo: new accssor z-interval specifies how often the Z file is updated
990 against the filesystem. We probably want no epoch stamp on these
991 entries. And we want to be able to filter the entries (e.g. no
992 by-modules and by-category tree)
994 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
996 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
997 into Z. We do not need the other files filled up with filesystem
1000 * need interface to query for a file in order to NOT call update on
1001 PAUSE a second time within a short time.
1003 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1005 * recommended update interval? Makes no sense, is different for
1012 change-log-default-name: "Todo"