1 2009-03-15 Andreas J. Koenig <andk@cpan.org>
3 * The bug below is -- after a lot of trying -- not reproducible on a
4 small script, only in the large test script. The closest to the output
10 my $x = "01237123229.8814";
15 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
17 $l = "1237123231.22458";
18 $r = "1237123231.22458";
22 Devel::Peek::Dump $x = $l <=> $r;
26 The checked in state at ... fails the test with my
27 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
28 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
29 in the test script always consistent.
31 * Todo: write a test that inserts a second dirty file with an already
34 * Bug in perl 5.10 on my 64bit box:
36 DB<98> Devel::Peek::Dump $l
37 SV = PVMG(0x19e0450) at 0x142a550
39 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
42 PV = 0x194ce70 "1237123231.22458"\0
46 DB<99> Devel::Peek::Dump $r
47 SV = PVMG(0x19e0240) at 0x142a3e8
49 FLAGS = (PADMY,POK,pPOK)
52 PV = 0x19ff900 "1237123231.22458"\0
56 DB<100> Devel::Peek::Dump $l <=> $r
57 SV = IV(0x19ea6e8) at 0x19ea6f0
59 FLAGS = (PADTMP,IOK,pIOK)
62 DB<101> 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<102> Devel::Peek::Dump $r
73 SV = PVMG(0x19e0240) at 0x142a3e8
75 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
78 PV = 0x19ff900 "1237123231.22458"\0
82 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
84 * Todo: reset "done" or "covered" and "minmax" after a dirty operation?
86 * currently recent_events has the side effect of setting dirtymark
87 because it forces a read on the file. That should be transparent, so
88 that the dirtymark call always forces a cache-able(?) read.
90 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
92 * $obj->merge ($other) needs to learn about equal epoch which may happen
93 since dirty_epoch intruded.
95 * Wontfix anytime soon: I think we currently do not support mkdir. Only
98 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
100 * Todo: continue working on update(...,$dirty_epoch). It must be
101 followed by a fast_aggregate!
103 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
105 * maybe we need a closest_entry or fitting_interval or something like
106 that. We want to merge an event into the middle of some recentfile.
107 First we do not know which file, then we do not know where to lock,
108 where to enter the new item, when and where to correct the dirtymark.
110 So my thought is we should first find which file.
112 Another part of my brain answers: what would happen if we would enter
113 the new file into the smallest file just like an ordinary new event,
114 just as an old event?
116 (1) we would write a duplicate timestamp? No, this would be easy to
119 (2) we would make the file large quickly? Yes, but so what? We are
120 changing the dirtymark, so are willing to disturb the downstream hosts.
122 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
124 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
128 https://rt.cpan.org/Ticket/Display.html?id=41199
130 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
132 Strangest output being HAKANARDO who managed to upload
134 Here is a better oneliner that includes also the first line of each
137 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
139 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
141 * memory leak in the syncher? It currently weighs 100M.
145 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
148 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
150 * lookup by epoch and by path and use this ability on the pause to never
151 again register a file twice that doesn't need it. Let's call it
154 * after the dirtymark is done: fill up recentfiles with fake (historic)
155 entries; fill up with individual corrections; algorithm maybe to be done
156 with bigfloat so that we can always place something in the middle
157 between two entries. Before we must switch to bigfloat we could try to
158 use Data::Float::nextup to get the.
160 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
163 * dirtymark now lives in Recentfile, needs to be used in rmirror.
165 * find out why the downloader died after a couple of hours without a net
166 connection. Write a test that survives the not-existence of the other
169 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
171 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
172 it makes less sense with tight coupling for secondary files. But asking
173 the server what the current epoch is might be cheaper on HTTP than on
174 rsync. (Needs to be evaluated)
176 * remove the 0.00 from the verbose overview in the Merged column in the
179 * write tests that expose the problems of the last few days: cascading
180 client/server roles, tight coupling for secondary RFs, deletes after
183 * Some day we might want to have policy options for the slave:
184 tight/loose/no coupling with upstream for secondary RFs. tight is what
185 we have now. loose would wait until a gap occurs that can be closed.
187 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
189 * revisit all $rfs->[$i+1] places if they now make sense still
191 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
193 * another bug is the fact that the mirror command deletes files before
194 it unhides the index file, thus confusing downstream slaves. We must not
195 delete before unhiding and must delete after unhiding. FIXED.
197 * new complication about the slave that is playing a server role.
198 Currently we mirror from newest to oldest with a hidden temporary file
199 as index. And when one file is finished, we unhide the index file.
200 Imagine the cascading server/slave is dead for a day. It then starts
201 mirroring again with the freshest thing and unhides the freshest index
202 file when it has worked through it. In that moment it exposes a time
203 hole. Because it now works on the second recentfile which is still
206 We currently do nothing special to converge after such a drop out. At
207 least not intentionally and robustly and thought through.
209 The algorithm we use to seed the next file needs quite a lot of more
210 robustness than it currently has. Something to do with looking at the
211 merged element of the next rf and when it has dropped off, we seed
212 immediately. And if it ramains dropped off, we seed again, of course.
214 Nope, looking from smaller to larger RFS we look at the merged element
215 of this RF and at the minmax/max element of the next RF. If that
216 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
219 And we need a public accessor seed and unseed or seeded. But not the mix
220 of public and private stuff that then is used behind the back.
222 And then the secondary* stuff must go.
224 And we must understand what the impact is on the DONE system. Can it go
225 unnoticed that there was a hole? And could the DONE system have decided
226 the hole is covered? This should be testable with three directories where
227 the middle stops working for a while. Done->merge is suspicious, we must
228 stop it from merging non-conflatable neighbors due to broken continuity.
232 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
234 * Slaven suggests to have the current epoch or the whole current
235 recentfile available from the HTTP server and take it away with
236 keepalive. This direction goes the granularity down to subseconds.
238 We might want to rewrite everything to factor out transport and allow
239 the whole thing to run via HTTP.
241 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
243 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
246 * smoker on k81 fetching from k75 to verify cascading works. See
247 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
249 * maybe the loop should wait for CHECKSUMS file after every upload. And
250 CPAN.pm needs to deal with timestamps in the future.
252 * do not forget the dirtymark!
254 Text: have a new flag on recentfiles with the meaning: if this
255 changes, you're required to run a full rsync over all the files. The
256 reason why we set it would probably be: some foul happened. we injected
257 files in arbitrary places or didn't inject them although they changed.
258 The content of the flag? Timestamp? The relation between the
259 recentfiles would have to be inheritance from the principal, because any
260 out of band changes would soon later propagate to the next recentfile.
262 By upping the flag often one can easily ruin the slaves.
264 last out of band change? dirtymark?
266 Anyway, this implies that we read a potentially existing recentfile
269 And it implies that we have an eventloop that keeps us busy in 2-3
270 cycles, one for current stuff (tight loop) and one for the recentfiles
271 (cascade when principal has changed), one for the old stuff after a
274 And it implies that the out-of-band change in any of the recentfiles
275 must have a lock on the principal file and there is the place to set the
278 * start a FAQ, especially quick start guide questions. Also to aid those
279 problematic areas where we have no good solution, like the "links"
282 * wish feedback when we are slow.
286 * Remove a few DEBUG statements.
288 * The multiple-rrr way of doing things needs a new option to rmirror,
289 like piecemeal or so. Not urgent because after the first pass through,
290 things run smoothely. It's only ugly during the first pass.
292 * I have the suspicion that the code is broken that decides if the
293 neighboring RF needs to be seeded. I fear when too much time has gone
294 between two calls (in our case more than one hour), it would not seed
295 the neighbor. Of course this will never be noticed, so we need a good
298 * local/localroot confusion: I currently pass both options but one must
301 * accounts for early birds on PAUSE rsync daemon.
303 * hardcoded 20 seconds
305 * who mirrors the index? DOING now.
307 * which CPAN mirrors offer rsync?
309 * visit all XXX, visit all _float places
311 * rename the pathdb stuff, it's too confusing. No idea how.
313 * rrr-inotify, backpan, rrr-register
315 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
317 * current bugs: the pathdb seems to get no reset, the seeding of the
318 secondaryttl stuff seems not to have an effect. Have helped myself with
319 a rand(10), need to fix this back. So not checked in. Does the rand
322 The rand thing helps. The secondaryttl stuff was in the wrong line,
325 The pathdb stuff was because I called either _pathdb or __pathdb on the
326 wrong object. FIXED now.
328 * It's not so beautiful if we never fetch the recentfiles that are not
329 the principal, even if this is correct behaviour. We really do not need
330 them after we have fetched the whole content.
332 OK, we want a switch for that: secondaryttl DONE
334 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
336 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
337 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
338 it lets 40 follow. FIXED
340 * See that the long running process really only updates the principal
341 file unless it has missed a timespan during which something happened. If
342 nothing happened, it must notice even when it misses the timespan. DONE
344 * we must throw away the pathdb when we have reached the end of Z. From
345 that moment we can have a very small pathdb because the only reason for
346 a pathdb is that we know to ignore old records in old files. We won't
347 need this pathdb again before the next full pass over the data is
348 necessary and then we will rebuild it as we go along. DONE
350 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
352 * I think, Done::register_one is doing wrong in that it does not
353 conflate neighboring pieces. The covered() method cannot do this because
354 it has no recent_events array at hand. But register_one has it and could
355 do it and for some reason misses to do it (sometimes).
357 This means that the three tests I just wrote can probably not survive
358 because they test with an already broken Done structure.
360 The art now is to detect how it happens, then to reproduce, then write a
363 So from the logfile this is what happens: we have a good interval with
364 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
365 top of it. Locally we now mirror F2 and open a new done interval for it.
366 Then we mirror F1 but this time with the timestamp T1b. And when we then
367 try to close the gap, we do not find T1 but instead something older. We
368 should gladly accept this older piece and this would fix this bug.
372 * bug to fix: when the 1h file changes while rmirror is running, we do
373 correctly sync the new files but never switch to the 6h file but rather
374 stay in a rather quick loop that fetches the 1h file again and again.
376 Is it possible that we initialize a new object? Or does
377 get_remote_recentfile_as_tempfile overwrite something in myself?
379 Want a new option: _runstatusfile => $file which frequently dumps the
380 state of all recentfiles to a file.
384 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
386 * Todo: now teach update to verify the timestamp is about to write
387 against the previous and use _increase_a_bit if it doesn't comply with
388 strict monotony. DONE
390 * The problem of rounding. So far perl's default precision was
391 sufficient. One day it won't be. FakeFloat has an easy job when it is
392 only reading and other machines have written correctly. But when we want
393 to write a floating point number that is a bit larger than the other
394 one, then we need our own idea of precision.
396 Slaven said: just append a "1". This might be going towards the end of
397 usability too quickly. I'd like something that actually uses the decimal
398 system. Well, appending a 1 also does this but...
400 E.g. we have 1.0. nextup on this architecture is starting with
401 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
402 taken the 1.0000000000000003 and the next user comes and the time tells
403 him 1.0 again. He has to beat my number without stepping over the
404 nextup. This is much less space than I had when I chose 1,2,3.
406 What is also irritating is that nextup is architecture dependent. The
407 128 bit guy must choose very long numbers to fit in between whereas the
408 other one with 16 bit uses larger steps. But then the algorithm is the
409 same for both, so that would be a nice thing.
411 I see two situation where we need this. One is when Time::HiRes returns
412 us a value that is <= the last entry in our recentfile. In this case
413 (let's call it the end-case) we must fill the region between that number
414 and the next higher native floating point number. The other is when we
415 inject an old file into an old recentfile (we would then also set a new
416 dirtymark). We find the integer value already taken and need a slightly
417 different one (let's call it the middle-case). The difference between
418 the two situations is that the next user will want to find something
419 higher than my number in the end-case and something lower than my number
422 So I suggest we give the function both a value and an upper bound and it
423 calculates us a primitive middle. The upper bound in the middle-case is
424 the next integer. The upper bound on the end-case is the nextup floating
425 point number. But the latter poses another problem: if we have occupied
426 the middle m between x and nextup(x), then the nextup(m) will probably
427 not be the same as nextup(x) because some rounding will take place
428 before the nextup is calculated and when the rounding reaches the
429 nextup(x), we will end up at nextup(nextup(x)).
431 So we really need to consider the nextup and the nextdown from there and
432 then the middle and that's the number we may approach asymptotically.
435 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
437 * consider deprecating the use of RECENT.recent as a symlink. It turns
438 out to need extra hoops with the rsync options and just isn't worth it.
439 Or maybe these extra hoops are needed anyway for the rest of the tree?
440 Nope, can't be the case because not all filesystems support symlinks.
442 But before doing the large step, I'll deprecate the call of
443 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
444 only introduced to resolve RECENT.recent and complicates the routine far
445 beyond what it deserves.
447 DONE. Won't deprecate RECENT.recent, just moved its handling to the
450 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
452 * I think it's a bug that the rsync_option links must be set to true in
453 order to support RECENT.recent and that nobody cares to set it
454 automatically. Similar for ignore_link_stat_errors.
456 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
458 * Todo: find all todos together and make a plan what is missing for a
461 - verifytree or something like that. fsck maybe.
463 - rersyncrecent, the script itself? What it do?
465 - a way to only mirror the recentfiles without mirroring the whole
466 remote system such that people can decide to mirror only partially see
467 also 2008-08-30. .shadow-xxx directory? this also needed for a
468 filesystem that is still incomplete and might need the mirrorfiles for
471 - long living objects that mirror again and again. Inject something
472 into ta, see how it goes over to tb.
474 - how do we continue filling up the DONE system when we use an object
475 for the second time? "fully covered" and "uptodate" or new terminology.
477 - overview called on the wrong file should be understandable
479 - the meta data field that must change when we fake something up so that
480 the downstream people know they have to re-fetch everything.
482 - how tolerant are we against missing files upstream? how do we keep
483 track? there are legitimate cases where we did read upstream index right
484 before a file got deleted there and then find that file as new and want
485 it. There are other cases that are not self healing and must be tracked
488 - how, exactly, do we have to deal with deletes? With rsync errors?
490 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
491 authors) failed: No such file or directory (2)
493 The file above is a delete in 1h and a new in file 1M and the
494 delete in the locally running rmirror did not get propagated to the 1M
495 object. Bug. And the consequence is a standstill.
497 It seems that a slave that works with a file below the principal needs
498 to merge things all the way up to get rid of later deletes. Or keep
499 track of all deletes and skip them later. So we need a trackdeletes.pm
500 similar to the done.pm?
502 see also 2008-08-20 about spurious deletes that really have no add
503 counterpart and yet they are not wrong.
505 - consider the effect when resyncing the recentfile takes longer than
506 the time per loop. Then we never rsync any file. We need to diagnose
507 that and force an increase of that loop time. But when we later are fast
508 enough again because the net has recovered, then we need to switch back
509 to original parameters. ERm, no, it's enough to keep syncing at least
510 one file before refetching an index file.
512 - remember to verify that no temp files are left lying around and the
515 - status file for not long running jobs that want to track upstream with
518 - revisit all XXX _float areas and study Sub::Exporter DONE
520 - persistent DB even though we just said we do not need it. Just for
521 extended capabilities and time savings when, for example, upstream
522 announces a reset and we get new recentfiles and could then limit
523 ourselves to a subset of files (those that have a changed epoch) in a
524 first pass and would only then do the loop to verify the rest. Or
527 * Todo: aggregate files should know their feed and finding the principal
528 should be done stepwise. (?)
530 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
531 left locks around. DONE
533 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
535 * maybe extend the _overview so that it always says if and where the
536 last file is in the next file and where the next event in the next rf
537 would lie. No, don't like this anymore. REJECT
539 * take the two new redundant tests out again, only the third must
542 * Todo: add a sanity check if the merged structure is really pointing to
543 a different rf and that this different rf is larger. DONE
545 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
547 * now test, if they are overlapping. And test if there is a file in the
548 next rf that would fit into this rf's interval.
550 1h 1222324012.8474 1222322541.7963 0.4086
551 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
552 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
553 1W 1222313218.3626 1221708477.5829 167.9835
555 I suspect that somebody writes a merged timestamp without having merged
556 and then somebody else relies on it.
558 If aggregate is running, the intervals must not be extravagated, if it
559 is not running, there must not be bounds, the total number of events in
560 the system must be counted and must be controlled throughout the tests.
561 That the test required the additional update was probably nonsense,
562 because aggregate can cut pieces too. FIXED & DONE
564 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
566 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
569 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
571 * Most apparent bug at the moment is that the recentfiles are fetched
572 too often. Only the principal should be fetched and if it has not
573 changed, the others should not be refetched. ATM I must admit that I'm
574 happy that we refetch more often than needed because I can more easily
575 fix bugs while the thing is running.
577 * Let's say, 1220474966.19501 is a timestamp of a file that is already
578 done but the done system does not know about it. The reason for the
579 failure is not known and we never reach the status uptodate because of
580 this. We must get over it.
582 Later it turns out that the origin server had a bug somewhere.
583 1220474966.19042 came after 1220474966.19501. Or better: it was in the
584 array of the recentfile one position above. The bug was my own.
586 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
588 * There is the race condition where the server does a delete and the
589 slave does not yet know and then tries to download it because he sees
590 the new. So for this time window we must be more tolerant against
591 failure. If we cannot download a file, we should just skip it and should
592 not retry immediately. The whole system should discover the lost thing
593 later. Keeping track with the DONE system should really be a no brainer.
595 But there is something more: the whole filesystem is a database and the
596 recentfiles are one possible representation of it. It's a pretty useful
597 representation I think that's why I have implemented something around
598 it. But for strictly local operation it has little value. For local
599 operation we would much rather have a database. So we would enter every
600 recentfile reading and every rsync operation and for every file the last
601 state change and what it leads to. Then we would always ignore older
602 records without the efforts involved with recentfiles.
604 The database would have: path,recentepoch,rsyncedon,deletedon
606 Oh well, not yet clear where this leads to.
608 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
610 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
611 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
613 * statusfile, maybe only on demand, alone to have a sharp debugging
614 tool. It is locked and all recentfiles dump themselves into it and we
615 can build a viewer that lets us know where we stand and what's inside.
617 * remember: only the principal recentfile needs expiration, all others
618 shall be expired by principal if it discovers that something has move
621 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
623 * Always check if we stringify to a higher value than in the entry
626 * And in covered make an additional check if we would be able to see a
627 numerical difference between the two numbers and if we can't then switch
628 to a different, more expensive algorithm. Do not want to be caught by
629 floating surprises. DONE
631 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
633 * caching has several aspects here: we can cache the interval of the
634 recentfile which only will change when the mtime of the file changes. We
635 must re-mirror the recentfile when its ttl has expired. Does have_read
636 tell you anything? It counts nothing at all. Only the mtime is
637 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
638 thing the have_mirrored because it is unrelated to the mtime.
640 * Robustness of floating point calculations! I always thought that the
641 string calculated by the origin server for the floating representation
642 of the epoch time is just a string. When we convert it to a number and
643 later back to a string, the other computer might come to a different
644 conclusion. This must not happen, we want to preserve it under any
645 circumstances. I will have to write tests with overlong sequences that
646 get lost in arithmetic and must see if all still works well. DONE
648 But one fragile point remains: if one host considers a>b and the other
649 one considers them == but no eq. To prevent this, we must probably do
650 some extra homework. DONE
652 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
654 * the concept of tracking DONE needs an object per recentfile that has
655 something like these methods:
657 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
659 covered() register() covered()
661 The unclear thing is how we translate points in time into intervals. We
662 could pass a reference to the current recent_events array when running
663 we_have(xxx) and let the DONE object iterate over it such that it only
664 has to store a list of intervals that can melt into each other. Ah, even
665 passing the list together with a list of indexes seems feasiable.
667 Or maybe ask for the inverted list?
669 Whenever the complete array is covered by the interval we say we are
670 fully covered and if the recentfile is not expired, we are uptodate.
672 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
674 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
676 * need a way to "return" the next entry after the end of a list. When
677 the caller says "before" or "after" we would like to know if he could
678 cover that interval/threshold or not because this influences the effect
679 of a newer timestamp of that recentfile. DONE with $opt{info}.
681 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
683 * one of the next things to tackle: the equivalent of csync2 -TIXU.
685 loop implies tixu (?). Nope, something like --statefile decides. Per
688 T test, I init, X including removals, U nodirtymark
690 So we have no concept of dirtymarks, we only trust that since we are
691 running we have observed everything steadily. But people will not let
692 this program run forever so we must consider both startup penalty and
693 book keeping for later runs. We keep this for later. For now we write a
694 long running mirror that merges several intervals.
696 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
698 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
701 * rersyncrecent, the script: default to one week. The name of the switch
702 is --after. Other switches? --loop!
704 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
706 * need a switch --skip-deletes (?)
708 * need a switch --enduser that tells us that the whole tempfile
709 discipline is not needed when there is no downstream user. (?)
711 Without this switch we cannot have a reasonable recent.pl that just
712 displays the recent additions. Either we accept to download everything.
713 Or we download temporary files without the typical rsync protocol
716 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
717 File::Temp::tempdir, this might be a win.
719 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
721 * apropos missing: we have no push, we never know the downstream
722 servers. People who know their downstream hosts and want to ascertain
723 something will want additional methods we have never thought about, like
724 update or delete a certain file.
726 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
728 * tempted to refactor rmirror into resolve_symlink, localize, etc.
729 Curious if rsync_options=links equal 0 vs. 1 will make the expected
732 * rsync options: it's a bit of a pain that we usually need several rsync
733 options, like compress, links, times, checksum and that there is no
734 reasonable default except the original rsync default. I think wee can
735 safely assume that the rsync options are shared between all recentfile
736 instances within one recent tree.
738 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
740 * deletes: if a delete follows an add quickly enough it may happen that
741 a downstream mirror did not see the add at all! It seems this needs to
742 be mentioned somewhere. The point here is that even if the downstream is
743 never missing the principal timeframe it may encounter a "delete" that
744 has no complimentary "add" anywhere.
746 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
748 * I suspect the treat of metadata is incorrect during read or something.
749 The bug that I am watching is that between 06:08 and 06:09 the 6h file
750 contained more than 6 hours worth of data. At 06:08 we merged into the
751 1d file. We need to take snapshots of the 6h file over the course of an
752 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
755 Much worse: watching the 1h file: right at the moment (at 06:35) it
756 covers 1218867584-1219120397 which is 70 hours.
758 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
759 08:19:44 2008, that is when I checked out last time, so it seems to be
760 aggregating and never truncating?
762 No, correct is: it is never truncating; but wrong is: it is aggregating.
763 It does receive a lot of events from time to time from a larger file.
764 Somehow a large file gets merged into the small one and because the
765 "meta/merged" attribute is missing, nobody is paying attention. I
766 believe that I can fix this by making sure that metadata are honoured
767 during read. DONE and test adjusted.
769 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
771 * grand renaming plan
773 remotebase => remoteroot to fit well with localroot DONE
774 local_path() => localroot seems to me should already work DONE
775 recentfile_basename => rfilename no need to stress it has no slash DONE
777 filenameroot??? Doesn't seem too bad to me today. Maybe something like
778 kern? It would anyway need a deprecation cycle because it is an
779 important constructor.
781 * I like the portability that Data::Serializer brings us but the price
782 is that some day we might find out that it is slowing us a bit. We'll
785 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
787 * should we not enter the interval of the principal (or the interval of
788 the merging file?) in every aggregated/merged file?
790 * we should aim at a first release and give up on thinking about
791 sanitizing stuff and zloop. Let's just admit that a full traditional
792 rsync is the only available sanitizer ATM. Otherwise it's complicated
793 stuff: sanitizing on the origin server, sanitizing on the slaves,
794 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
795 get the basics out before this becomes a major cause for mess.
797 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
799 * On OSes not supporting symlinks we expect that RECENT.recent contains
800 the contents of the principal recentfile. Actually this is identical on
801 systems supporting symlinks. Simple, what follows from that is that we
802 need to keep the serializer in the metadata because we cannot read it
803 from the filename, doesn't it? Of course not. It's a chicken and egg
804 problem. This leaves us with the problem to actually parse the
805 serialized data to find out in which format it is. So who can do the 4
806 or 5 magics we wanted to support? File::LibMagic?
808 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
810 * remotebase and recentfile_basename are ugly names. Now that we need a
811 word for the shortest/principal/driving recentfile too we should do
814 localroot is good. rfile is good. local_path() is bad, local_path($path)
815 is medium, filenameroot() is bad, remotebase is bad, recentfile is
818 Up to now remotebase was the string that described the remote root
819 directory in rsync notation, like pause.perl.org::authors. And
820 recentfile_basename was "RECENT-1h.yaml".
822 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
824 * The test that was added in today's checkin is a good start for a test
825 of rmirror. We should have more methods in Recent.pm: verify,
826 addmissingfiles. We should verify the current tree, then rmirror it and
827 then verifytree the copy. We could then add some arbitrary file and let
828 it be discovered by addmissingfiles, then rmirror again and then
829 verifytree the copy again.
831 Then we could start stealing from csync2 sqlite database [no port to
832 OSX!] and fill a local DB. And methods to compare the database with the
833 recentfiles. Our strength is that in principle we could maintain state
834 with a single float. We have synced up to 1234567890.123456. If the Z
835 file does not add new files all we have to do is mirror the new ones and
838 This makes it clear that we should extend current protocol and declare
839 that we cheat when we add files too late, just to help the other end
840 keeping track. Ah yes, that's what was meant when zloop was mentioned
843 Maybe need to revisit File::Mirror to help me with this task.
845 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
847 * There must be an allow-me-to-truncate flag in every recentfile.
848 Without it one could construct a sequence of updates winning the locking
849 battle against the aggregator. Only if an aggregator has managed to
850 merge data over to the next level, truncating can be allowed. DONE with
853 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
855 * We should probably guarantee that no duplicates enter the aggregator
858 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
860 * To get merge operation faster would need a good benchmark test. What
861 02 spits out isn't reliable enough and is dominated by many other
864 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
865 Author: Andreas J. Koenig <k@k75.(none)>
866 Date: Sat Aug 2 07:58:04 2008 +0200
870 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
871 Author: Andreas J. Koenig <k@k75.(none)>
872 Date: Sat Aug 2 11:40:29 2008 +0200
874 there was a lot of trying but the effect is hardly measurable with
877 * overhead of connecting seems high. When setting
878 max_files_per_connection to 1 we see that.
880 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
882 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
884 25138.2215600014/3600 = 6.98283932222261
886 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
887 be a bug in the aggregator. FIXED
889 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
891 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
892 we have it, can we mirror it, mirror it!
894 I fear this needs a new class which might be called
895 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
896 and treat them as an entity. I realize that a single recentfile may be
897 sufficient for certain tasks and that it is handy for the low level
898 programmer but it is not nice to use. If there is a delete in the 1h
899 file then the 6h file still contains it. Seekers of the best information
900 need to combine at least some of the recentfiles most of the time.
902 There is the place for the Z loop!
904 But the combination is something to collect in a database, isn't it. Did
905 csync2 just harrumph?
907 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
909 * it just occurred to me that hosts in the same mirroring pool could
910 help out each other even without rewriting the recentfile. Just fetch
911 the stuff to mirror from several places, bingo. But that's something
912 that should rather live in a separate package or in rsync directly.
914 * cronjobs are unsuited because with ntp they would all come at the full
915 minute and disturb each other. Besides that I'd hate to have a backbone
916 with more than a few seconds latency.
918 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
920 * a second rsync server with access control for PAUSE. Port? 873 is the
921 standard port, let's take 8873.
923 * if there were a filesystem based on this, it would have a slow access
924 to inexistent files. It would probably provide wrong readdir (only based
925 on current content) or also a slow one (based on a recentfile written
926 after the call). But it would provide fast access to existing files. Or
927 one would deliberately allow slightly blurred answers based on some
928 sqlite reflection of the recentfiles.
930 * todo: write a variant of mirror() that combines two or more
931 recentfiles and treats them like one
933 * todo: signal handler to remove the tempfile
935 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
937 * now that we have the symlink I forgot how it should be used in
940 * the z loop: add missing files to Z file. Just append them (instead of
941 prepending). So one guy prepends something from the Y file from time to
942 time and another guy appends something rather frequently. Collecting
943 pond. When Y merges into Z, things get epoch and the collecting pond
944 gets smaller. What exactly are "missing files"?
946 take note of current epoch of the alpha file, let's call it the
949 find all files on disk
951 remove all files registered in the recentworld up to recent-ts
953 remove all files that have been deleted after recent-ts according to
956 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
958 * rersyncrecent might be a cronjob with a (locked) state file which
959 contains things like after and maybe last z sync or such?
961 rrr-mirror might be an alternative name but how would we justify the
962 three Rs when there is no Re-Rsync-Recent?
964 With the --loop parameter it is an endless loop, without it is no loop.
965 At least this is simple.
967 * todo: new accssor z-interval specifies how often the Z file is updated
968 against the filesystem. We probably want no epoch stamp on these
969 entries. And we want to be able to filter the entries (e.g. no
970 by-modules and by-category tree)
972 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
974 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
975 into Z. We do not need the other files filled up with filesystem
978 * need interface to query for a file in order to NOT call update on
979 PAUSE a second time within a short time.
981 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
983 * recommended update interval? Makes no sense, is different for
990 change-log-default-name: "Todo"