1 2009-03-17 Andreas J. Koenig <andk@cpan.org>
3 * todo: integrate a dirty update with two aggregate calls before
4 unlocking for frictionless dirtying
6 * todo: start the second rsync daemon on pause
8 * todo: move index files to .recent
10 2009-03-16 Andreas J. Koenig <andk@cpan.org>
12 * What was the resolution of the mirror.pl delete hook bug? Do we call
13 the delete hook when pause removes a file from MUIR?
15 * Today on pause: Updating 2a13fba..29f284d and installing it for
16 /usr/local/perl-5.10.0{,-RC2}
18 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
19 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
21 2009-03-15 Andreas J. Koenig <andk@cpan.org>
23 * currently recent_events has the side effect of setting dirtymark
24 because it forces a read on the file. That should be transparent, so
25 that the dirtymark call always forces a cache-able(?) read.
27 * The bug below is -- after a lot of trying -- not reproducible on a
28 small script, only in the large test script. The closest to the output
34 my $x = "01237123229.8814";
39 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
41 $l = "1237123231.22458";
42 $r = "1237123231.22458";
46 Devel::Peek::Dump $x = $l <=> $r;
50 The checked in state at c404a85 fails the test with my
51 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
52 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
53 in the test script always consistent.
55 * Todo: write a test that inserts a second dirty file with an already
56 existing timestamp. DONE
58 * Bug in perl 5.10 on my 64bit box:
60 DB<98> Devel::Peek::Dump $l
61 SV = PVMG(0x19e0450) at 0x142a550
63 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
66 PV = 0x194ce70 "1237123231.22458"\0
70 DB<99> Devel::Peek::Dump $r
71 SV = PVMG(0x19e0240) at 0x142a3e8
73 FLAGS = (PADMY,POK,pPOK)
76 PV = 0x19ff900 "1237123231.22458"\0
80 DB<100> Devel::Peek::Dump $l <=> $r
81 SV = IV(0x19ea6e8) at 0x19ea6f0
83 FLAGS = (PADTMP,IOK,pIOK)
86 DB<101> Devel::Peek::Dump $l
87 SV = PVMG(0x19e0450) at 0x142a550
89 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
92 PV = 0x194ce70 "1237123231.22458"\0
96 DB<102> Devel::Peek::Dump $r
97 SV = PVMG(0x19e0240) at 0x142a3e8
99 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
101 NV = 1237123231.22458
102 PV = 0x19ff900 "1237123231.22458"\0
106 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
108 * Todo: reset "done" or "covered" and "minmax" after a dirty operation?
110 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
112 * $obj->merge ($other) needs to learn about equal epoch which may happen
113 since dirty_epoch intruded.
115 * Wontfix anytime soon: I think we currently do not support mkdir. Only
118 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
120 * Todo: continue working on update(...,$dirty_epoch). It must be
121 followed by a fast_aggregate!
123 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
125 * maybe we need a closest_entry or fitting_interval or something like
126 that. We want to merge an event into the middle of some recentfile.
127 First we do not know which file, then we do not know where to lock,
128 where to enter the new item, when and where to correct the dirtymark.
130 So my thought is we should first find which file.
132 Another part of my brain answers: what would happen if we would enter
133 the new file into the smallest file just like an ordinary new event,
134 just as an old event?
136 (1) we would write a duplicate timestamp? No, this would be easy to
139 (2) we would make the file large quickly? Yes, but so what? We are
140 changing the dirtymark, so are willing to disturb the downstream hosts.
142 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
144 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
148 https://rt.cpan.org/Ticket/Display.html?id=41199
150 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
152 Strangest output being HAKANARDO who managed to upload
154 Here is a better oneliner that includes also the first line of each
157 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
159 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
161 * memory leak in the syncher? It currently weighs 100M.
165 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
168 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
170 * lookup by epoch and by path and use this ability on the pause to never
171 again register a file twice that doesn't need it. Let's call it
174 * after the dirtymark is done: fill up recentfiles with fake (historic)
175 entries; fill up with individual corrections; algorithm maybe to be done
176 with bigfloat so that we can always place something in the middle
177 between two entries. Before we must switch to bigfloat we could try to
178 use Data::Float::nextup to get the.
180 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
183 * dirtymark now lives in Recentfile, needs to be used in rmirror.
185 * find out why the downloader died after a couple of hours without a net
186 connection. Write a test that survives the not-existence of the other
189 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
191 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
192 it makes less sense with tight coupling for secondary files. But asking
193 the server what the current epoch is might be cheaper on HTTP than on
194 rsync. (Needs to be evaluated)
196 * remove the 0.00 from the verbose overview in the Merged column in the
199 * write tests that expose the problems of the last few days: cascading
200 client/server roles, tight coupling for secondary RFs, deletes after
203 * Some day we might want to have policy options for the slave:
204 tight/loose/no coupling with upstream for secondary RFs. tight is what
205 we have now. loose would wait until a gap occurs that can be closed.
207 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
209 * revisit all $rfs->[$i+1] places if they now make sense still
211 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
213 * another bug is the fact that the mirror command deletes files before
214 it unhides the index file, thus confusing downstream slaves. We must not
215 delete before unhiding and must delete after unhiding. FIXED.
217 * new complication about the slave that is playing a server role.
218 Currently we mirror from newest to oldest with a hidden temporary file
219 as index. And when one file is finished, we unhide the index file.
220 Imagine the cascading server/slave is dead for a day. It then starts
221 mirroring again with the freshest thing and unhides the freshest index
222 file when it has worked through it. In that moment it exposes a time
223 hole. Because it now works on the second recentfile which is still
226 We currently do nothing special to converge after such a drop out. At
227 least not intentionally and robustly and thought through.
229 The algorithm we use to seed the next file needs quite a lot of more
230 robustness than it currently has. Something to do with looking at the
231 merged element of the next rf and when it has dropped off, we seed
232 immediately. And if it ramains dropped off, we seed again, of course.
234 Nope, looking from smaller to larger RFS we look at the merged element
235 of this RF and at the minmax/max element of the next RF. If that
236 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
239 And we need a public accessor seed and unseed or seeded. But not the mix
240 of public and private stuff that then is used behind the back.
242 And then the secondary* stuff must go.
244 And we must understand what the impact is on the DONE system. Can it go
245 unnoticed that there was a hole? And could the DONE system have decided
246 the hole is covered? This should be testable with three directories where
247 the middle stops working for a while. Done->merge is suspicious, we must
248 stop it from merging non-conflatable neighbors due to broken continuity.
252 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
254 * Slaven suggests to have the current epoch or the whole current
255 recentfile available from the HTTP server and take it away with
256 keepalive. This direction goes the granularity down to subseconds.
258 We might want to rewrite everything to factor out transport and allow
259 the whole thing to run via HTTP.
261 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
263 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
266 * smoker on k81 fetching from k75 to verify cascading works. See
267 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
269 * maybe the loop should wait for CHECKSUMS file after every upload. And
270 CPAN.pm needs to deal with timestamps in the future.
272 * do not forget the dirtymark!
274 Text: have a new flag on recentfiles with the meaning: if this
275 changes, you're required to run a full rsync over all the files. The
276 reason why we set it would probably be: some foul happened. we injected
277 files in arbitrary places or didn't inject them although they changed.
278 The content of the flag? Timestamp? The relation between the
279 recentfiles would have to be inheritance from the principal, because any
280 out of band changes would soon later propagate to the next recentfile.
282 By upping the flag often one can easily ruin the slaves.
284 last out of band change? dirtymark?
286 Anyway, this implies that we read a potentially existing recentfile
289 And it implies that we have an eventloop that keeps us busy in 2-3
290 cycles, one for current stuff (tight loop) and one for the recentfiles
291 (cascade when principal has changed), one for the old stuff after a
294 And it implies that the out-of-band change in any of the recentfiles
295 must have a lock on the principal file and there is the place to set the
298 * start a FAQ, especially quick start guide questions. Also to aid those
299 problematic areas where we have no good solution, like the "links"
302 * wish feedback when we are slow.
306 * Remove a few DEBUG statements.
308 * The multiple-rrr way of doing things needs a new option to rmirror,
309 like piecemeal or so. Not urgent because after the first pass through,
310 things run smoothely. It's only ugly during the first pass.
312 * I have the suspicion that the code is broken that decides if the
313 neighboring RF needs to be seeded. I fear when too much time has gone
314 between two calls (in our case more than one hour), it would not seed
315 the neighbor. Of course this will never be noticed, so we need a good
318 * local/localroot confusion: I currently pass both options but one must
321 * accounts for early birds on PAUSE rsync daemon.
323 * hardcoded 20 seconds
325 * who mirrors the index? DOING now.
327 * which CPAN mirrors offer rsync?
329 * visit all XXX, visit all _float places
331 * rename the pathdb stuff, it's too confusing. No idea how.
333 * rrr-inotify, backpan, rrr-register
335 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
337 * current bugs: the pathdb seems to get no reset, the seeding of the
338 secondaryttl stuff seems not to have an effect. Have helped myself with
339 a rand(10), need to fix this back. So not checked in. Does the rand
342 The rand thing helps. The secondaryttl stuff was in the wrong line,
345 The pathdb stuff was because I called either _pathdb or __pathdb on the
346 wrong object. FIXED now.
348 * It's not so beautiful if we never fetch the recentfiles that are not
349 the principal, even if this is correct behaviour. We really do not need
350 them after we have fetched the whole content.
352 OK, we want a switch for that: secondaryttl DONE
354 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
356 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
357 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
358 it lets 40 follow. FIXED
360 * See that the long running process really only updates the principal
361 file unless it has missed a timespan during which something happened. If
362 nothing happened, it must notice even when it misses the timespan. DONE
364 * we must throw away the pathdb when we have reached the end of Z. From
365 that moment we can have a very small pathdb because the only reason for
366 a pathdb is that we know to ignore old records in old files. We won't
367 need this pathdb again before the next full pass over the data is
368 necessary and then we will rebuild it as we go along. DONE
370 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
372 * I think, Done::register_one is doing wrong in that it does not
373 conflate neighboring pieces. The covered() method cannot do this because
374 it has no recent_events array at hand. But register_one has it and could
375 do it and for some reason misses to do it (sometimes).
377 This means that the three tests I just wrote can probably not survive
378 because they test with an already broken Done structure.
380 The art now is to detect how it happens, then to reproduce, then write a
383 So from the logfile this is what happens: we have a good interval with
384 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
385 top of it. Locally we now mirror F2 and open a new done interval for it.
386 Then we mirror F1 but this time with the timestamp T1b. And when we then
387 try to close the gap, we do not find T1 but instead something older. We
388 should gladly accept this older piece and this would fix this bug.
392 * bug to fix: when the 1h file changes while rmirror is running, we do
393 correctly sync the new files but never switch to the 6h file but rather
394 stay in a rather quick loop that fetches the 1h file again and again.
396 Is it possible that we initialize a new object? Or does
397 get_remote_recentfile_as_tempfile overwrite something in myself?
399 Want a new option: _runstatusfile => $file which frequently dumps the
400 state of all recentfiles to a file.
404 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
406 * Todo: now teach update to verify the timestamp is about to write
407 against the previous and use _increase_a_bit if it doesn't comply with
408 strict monotony. DONE
410 * The problem of rounding. So far perl's default precision was
411 sufficient. One day it won't be. FakeFloat has an easy job when it is
412 only reading and other machines have written correctly. But when we want
413 to write a floating point number that is a bit larger than the other
414 one, then we need our own idea of precision.
416 Slaven said: just append a "1". This might be going towards the end of
417 usability too quickly. I'd like something that actually uses the decimal
418 system. Well, appending a 1 also does this but...
420 E.g. we have 1.0. nextup on this architecture is starting with
421 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
422 taken the 1.0000000000000003 and the next user comes and the time tells
423 him 1.0 again. He has to beat my number without stepping over the
424 nextup. This is much less space than I had when I chose 1,2,3.
426 What is also irritating is that nextup is architecture dependent. The
427 128 bit guy must choose very long numbers to fit in between whereas the
428 other one with 16 bit uses larger steps. But then the algorithm is the
429 same for both, so that would be a nice thing.
431 I see two situation where we need this. One is when Time::HiRes returns
432 us a value that is <= the last entry in our recentfile. In this case
433 (let's call it the end-case) we must fill the region between that number
434 and the next higher native floating point number. The other is when we
435 inject an old file into an old recentfile (we would then also set a new
436 dirtymark). We find the integer value already taken and need a slightly
437 different one (let's call it the middle-case). The difference between
438 the two situations is that the next user will want to find something
439 higher than my number in the end-case and something lower than my number
442 So I suggest we give the function both a value and an upper bound and it
443 calculates us a primitive middle. The upper bound in the middle-case is
444 the next integer. The upper bound on the end-case is the nextup floating
445 point number. But the latter poses another problem: if we have occupied
446 the middle m between x and nextup(x), then the nextup(m) will probably
447 not be the same as nextup(x) because some rounding will take place
448 before the nextup is calculated and when the rounding reaches the
449 nextup(x), we will end up at nextup(nextup(x)).
451 So we really need to consider the nextup and the nextdown from there and
452 then the middle and that's the number we may approach asymptotically.
455 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
457 * consider deprecating the use of RECENT.recent as a symlink. It turns
458 out to need extra hoops with the rsync options and just isn't worth it.
459 Or maybe these extra hoops are needed anyway for the rest of the tree?
460 Nope, can't be the case because not all filesystems support symlinks.
462 But before doing the large step, I'll deprecate the call of
463 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
464 only introduced to resolve RECENT.recent and complicates the routine far
465 beyond what it deserves.
467 DONE. Won't deprecate RECENT.recent, just moved its handling to the
470 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
472 * I think it's a bug that the rsync_option links must be set to true in
473 order to support RECENT.recent and that nobody cares to set it
474 automatically. Similar for ignore_link_stat_errors.
476 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
478 * Todo: find all todos together and make a plan what is missing for a
481 - verifytree or something like that. fsck maybe.
483 - rersyncrecent, the script itself? What it do?
485 - a way to only mirror the recentfiles without mirroring the whole
486 remote system such that people can decide to mirror only partially see
487 also 2008-08-30. .shadow-xxx directory? this also needed for a
488 filesystem that is still incomplete and might need the mirrorfiles for
491 - long living objects that mirror again and again. Inject something
492 into ta, see how it goes over to tb.
494 - how do we continue filling up the DONE system when we use an object
495 for the second time? "fully covered" and "uptodate" or new terminology.
497 - overview called on the wrong file should be understandable
499 - the meta data field that must change when we fake something up so that
500 the downstream people know they have to re-fetch everything.
502 - how tolerant are we against missing files upstream? how do we keep
503 track? there are legitimate cases where we did read upstream index right
504 before a file got deleted there and then find that file as new and want
505 it. There are other cases that are not self healing and must be tracked
508 - how, exactly, do we have to deal with deletes? With rsync errors?
510 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
511 authors) failed: No such file or directory (2)
513 The file above is a delete in 1h and a new in file 1M and the
514 delete in the locally running rmirror did not get propagated to the 1M
515 object. Bug. And the consequence is a standstill.
517 It seems that a slave that works with a file below the principal needs
518 to merge things all the way up to get rid of later deletes. Or keep
519 track of all deletes and skip them later. So we need a trackdeletes.pm
520 similar to the done.pm?
522 see also 2008-08-20 about spurious deletes that really have no add
523 counterpart and yet they are not wrong.
525 - consider the effect when resyncing the recentfile takes longer than
526 the time per loop. Then we never rsync any file. We need to diagnose
527 that and force an increase of that loop time. But when we later are fast
528 enough again because the net has recovered, then we need to switch back
529 to original parameters. ERm, no, it's enough to keep syncing at least
530 one file before refetching an index file.
532 - remember to verify that no temp files are left lying around and the
535 - status file for not long running jobs that want to track upstream with
538 - revisit all XXX _float areas and study Sub::Exporter DONE
540 - persistent DB even though we just said we do not need it. Just for
541 extended capabilities and time savings when, for example, upstream
542 announces a reset and we get new recentfiles and could then limit
543 ourselves to a subset of files (those that have a changed epoch) in a
544 first pass and would only then do the loop to verify the rest. Or
547 * Todo: aggregate files should know their feed and finding the principal
548 should be done stepwise. (?)
550 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
551 left locks around. DONE
553 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
555 * maybe extend the _overview so that it always says if and where the
556 last file is in the next file and where the next event in the next rf
557 would lie. No, don't like this anymore. REJECT
559 * take the two new redundant tests out again, only the third must
562 * Todo: add a sanity check if the merged structure is really pointing to
563 a different rf and that this different rf is larger. DONE
565 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
567 * now test, if they are overlapping. And test if there is a file in the
568 next rf that would fit into this rf's interval.
570 1h 1222324012.8474 1222322541.7963 0.4086
571 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
572 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
573 1W 1222313218.3626 1221708477.5829 167.9835
575 I suspect that somebody writes a merged timestamp without having merged
576 and then somebody else relies on it.
578 If aggregate is running, the intervals must not be extravagated, if it
579 is not running, there must not be bounds, the total number of events in
580 the system must be counted and must be controlled throughout the tests.
581 That the test required the additional update was probably nonsense,
582 because aggregate can cut pieces too. FIXED & DONE
584 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
586 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
589 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
591 * Most apparent bug at the moment is that the recentfiles are fetched
592 too often. Only the principal should be fetched and if it has not
593 changed, the others should not be refetched. ATM I must admit that I'm
594 happy that we refetch more often than needed because I can more easily
595 fix bugs while the thing is running.
597 * Let's say, 1220474966.19501 is a timestamp of a file that is already
598 done but the done system does not know about it. The reason for the
599 failure is not known and we never reach the status uptodate because of
600 this. We must get over it.
602 Later it turns out that the origin server had a bug somewhere.
603 1220474966.19042 came after 1220474966.19501. Or better: it was in the
604 array of the recentfile one position above. The bug was my own.
606 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
608 * There is the race condition where the server does a delete and the
609 slave does not yet know and then tries to download it because he sees
610 the new. So for this time window we must be more tolerant against
611 failure. If we cannot download a file, we should just skip it and should
612 not retry immediately. The whole system should discover the lost thing
613 later. Keeping track with the DONE system should really be a no brainer.
615 But there is something more: the whole filesystem is a database and the
616 recentfiles are one possible representation of it. It's a pretty useful
617 representation I think that's why I have implemented something around
618 it. But for strictly local operation it has little value. For local
619 operation we would much rather have a database. So we would enter every
620 recentfile reading and every rsync operation and for every file the last
621 state change and what it leads to. Then we would always ignore older
622 records without the efforts involved with recentfiles.
624 The database would have: path,recentepoch,rsyncedon,deletedon
626 Oh well, not yet clear where this leads to.
628 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
630 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
631 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
633 * statusfile, maybe only on demand, alone to have a sharp debugging
634 tool. It is locked and all recentfiles dump themselves into it and we
635 can build a viewer that lets us know where we stand and what's inside.
637 * remember: only the principal recentfile needs expiration, all others
638 shall be expired by principal if it discovers that something has move
641 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
643 * Always check if we stringify to a higher value than in the entry
646 * And in covered make an additional check if we would be able to see a
647 numerical difference between the two numbers and if we can't then switch
648 to a different, more expensive algorithm. Do not want to be caught by
649 floating surprises. DONE
651 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
653 * caching has several aspects here: we can cache the interval of the
654 recentfile which only will change when the mtime of the file changes. We
655 must re-mirror the recentfile when its ttl has expired. Does have_read
656 tell you anything? It counts nothing at all. Only the mtime is
657 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
658 thing the have_mirrored because it is unrelated to the mtime.
660 * Robustness of floating point calculations! I always thought that the
661 string calculated by the origin server for the floating representation
662 of the epoch time is just a string. When we convert it to a number and
663 later back to a string, the other computer might come to a different
664 conclusion. This must not happen, we want to preserve it under any
665 circumstances. I will have to write tests with overlong sequences that
666 get lost in arithmetic and must see if all still works well. DONE
668 But one fragile point remains: if one host considers a>b and the other
669 one considers them == but no eq. To prevent this, we must probably do
670 some extra homework. DONE
672 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
674 * the concept of tracking DONE needs an object per recentfile that has
675 something like these methods:
677 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
679 covered() register() covered()
681 The unclear thing is how we translate points in time into intervals. We
682 could pass a reference to the current recent_events array when running
683 we_have(xxx) and let the DONE object iterate over it such that it only
684 has to store a list of intervals that can melt into each other. Ah, even
685 passing the list together with a list of indexes seems feasiable.
687 Or maybe ask for the inverted list?
689 Whenever the complete array is covered by the interval we say we are
690 fully covered and if the recentfile is not expired, we are uptodate.
692 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
694 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
696 * need a way to "return" the next entry after the end of a list. When
697 the caller says "before" or "after" we would like to know if he could
698 cover that interval/threshold or not because this influences the effect
699 of a newer timestamp of that recentfile. DONE with $opt{info}.
701 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
703 * one of the next things to tackle: the equivalent of csync2 -TIXU.
705 loop implies tixu (?). Nope, something like --statefile decides. Per
708 T test, I init, X including removals, U nodirtymark
710 So we have no concept of dirtymarks, we only trust that since we are
711 running we have observed everything steadily. But people will not let
712 this program run forever so we must consider both startup penalty and
713 book keeping for later runs. We keep this for later. For now we write a
714 long running mirror that merges several intervals.
716 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
718 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
721 * rersyncrecent, the script: default to one week. The name of the switch
722 is --after. Other switches? --loop!
724 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
726 * need a switch --skip-deletes (?)
728 * need a switch --enduser that tells us that the whole tempfile
729 discipline is not needed when there is no downstream user. (?)
731 Without this switch we cannot have a reasonable recent.pl that just
732 displays the recent additions. Either we accept to download everything.
733 Or we download temporary files without the typical rsync protocol
736 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
737 File::Temp::tempdir, this might be a win.
739 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
741 * apropos missing: we have no push, we never know the downstream
742 servers. People who know their downstream hosts and want to ascertain
743 something will want additional methods we have never thought about, like
744 update or delete a certain file.
746 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
748 * tempted to refactor rmirror into resolve_symlink, localize, etc.
749 Curious if rsync_options=links equal 0 vs. 1 will make the expected
752 * rsync options: it's a bit of a pain that we usually need several rsync
753 options, like compress, links, times, checksum and that there is no
754 reasonable default except the original rsync default. I think wee can
755 safely assume that the rsync options are shared between all recentfile
756 instances within one recent tree.
758 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
760 * deletes: if a delete follows an add quickly enough it may happen that
761 a downstream mirror did not see the add at all! It seems this needs to
762 be mentioned somewhere. The point here is that even if the downstream is
763 never missing the principal timeframe it may encounter a "delete" that
764 has no complimentary "add" anywhere.
766 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
768 * I suspect the treat of metadata is incorrect during read or something.
769 The bug that I am watching is that between 06:08 and 06:09 the 6h file
770 contained more than 6 hours worth of data. At 06:08 we merged into the
771 1d file. We need to take snapshots of the 6h file over the course of an
772 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
775 Much worse: watching the 1h file: right at the moment (at 06:35) it
776 covers 1218867584-1219120397 which is 70 hours.
778 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
779 08:19:44 2008, that is when I checked out last time, so it seems to be
780 aggregating and never truncating?
782 No, correct is: it is never truncating; but wrong is: it is aggregating.
783 It does receive a lot of events from time to time from a larger file.
784 Somehow a large file gets merged into the small one and because the
785 "meta/merged" attribute is missing, nobody is paying attention. I
786 believe that I can fix this by making sure that metadata are honoured
787 during read. DONE and test adjusted.
789 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
791 * grand renaming plan
793 remotebase => remoteroot to fit well with localroot DONE
794 local_path() => localroot seems to me should already work DONE
795 recentfile_basename => rfilename no need to stress it has no slash DONE
797 filenameroot??? Doesn't seem too bad to me today. Maybe something like
798 kern? It would anyway need a deprecation cycle because it is an
799 important constructor.
801 * I like the portability that Data::Serializer brings us but the price
802 is that some day we might find out that it is slowing us a bit. We'll
805 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
807 * should we not enter the interval of the principal (or the interval of
808 the merging file?) in every aggregated/merged file?
810 * we should aim at a first release and give up on thinking about
811 sanitizing stuff and zloop. Let's just admit that a full traditional
812 rsync is the only available sanitizer ATM. Otherwise it's complicated
813 stuff: sanitizing on the origin server, sanitizing on the slaves,
814 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
815 get the basics out before this becomes a major cause for mess.
817 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
819 * On OSes not supporting symlinks we expect that RECENT.recent contains
820 the contents of the principal recentfile. Actually this is identical on
821 systems supporting symlinks. Simple, what follows from that is that we
822 need to keep the serializer in the metadata because we cannot read it
823 from the filename, doesn't it? Of course not. It's a chicken and egg
824 problem. This leaves us with the problem to actually parse the
825 serialized data to find out in which format it is. So who can do the 4
826 or 5 magics we wanted to support? File::LibMagic?
828 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
830 * remotebase and recentfile_basename are ugly names. Now that we need a
831 word for the shortest/principal/driving recentfile too we should do
834 localroot is good. rfile is good. local_path() is bad, local_path($path)
835 is medium, filenameroot() is bad, remotebase is bad, recentfile is
838 Up to now remotebase was the string that described the remote root
839 directory in rsync notation, like pause.perl.org::authors. And
840 recentfile_basename was "RECENT-1h.yaml".
842 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
844 * The test that was added in today's checkin is a good start for a test
845 of rmirror. We should have more methods in Recent.pm: verify,
846 addmissingfiles. We should verify the current tree, then rmirror it and
847 then verifytree the copy. We could then add some arbitrary file and let
848 it be discovered by addmissingfiles, then rmirror again and then
849 verifytree the copy again.
851 Then we could start stealing from csync2 sqlite database [no port to
852 OSX!] and fill a local DB. And methods to compare the database with the
853 recentfiles. Our strength is that in principle we could maintain state
854 with a single float. We have synced up to 1234567890.123456. If the Z
855 file does not add new files all we have to do is mirror the new ones and
858 This makes it clear that we should extend current protocol and declare
859 that we cheat when we add files too late, just to help the other end
860 keeping track. Ah yes, that's what was meant when zloop was mentioned
863 Maybe need to revisit File::Mirror to help me with this task.
865 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
867 * There must be an allow-me-to-truncate flag in every recentfile.
868 Without it one could construct a sequence of updates winning the locking
869 battle against the aggregator. Only if an aggregator has managed to
870 merge data over to the next level, truncating can be allowed. DONE with
873 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
875 * We should probably guarantee that no duplicates enter the aggregator
878 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
880 * To get merge operation faster would need a good benchmark test. What
881 02 spits out isn't reliable enough and is dominated by many other
884 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
885 Author: Andreas J. Koenig <k@k75.(none)>
886 Date: Sat Aug 2 07:58:04 2008 +0200
890 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
891 Author: Andreas J. Koenig <k@k75.(none)>
892 Date: Sat Aug 2 11:40:29 2008 +0200
894 there was a lot of trying but the effect is hardly measurable with
897 * overhead of connecting seems high. When setting
898 max_files_per_connection to 1 we see that.
900 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
902 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
904 25138.2215600014/3600 = 6.98283932222261
906 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
907 be a bug in the aggregator. FIXED
909 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
911 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
912 we have it, can we mirror it, mirror it!
914 I fear this needs a new class which might be called
915 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
916 and treat them as an entity. I realize that a single recentfile may be
917 sufficient for certain tasks and that it is handy for the low level
918 programmer but it is not nice to use. If there is a delete in the 1h
919 file then the 6h file still contains it. Seekers of the best information
920 need to combine at least some of the recentfiles most of the time.
922 There is the place for the Z loop!
924 But the combination is something to collect in a database, isn't it. Did
925 csync2 just harrumph?
927 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
929 * it just occurred to me that hosts in the same mirroring pool could
930 help out each other even without rewriting the recentfile. Just fetch
931 the stuff to mirror from several places, bingo. But that's something
932 that should rather live in a separate package or in rsync directly.
934 * cronjobs are unsuited because with ntp they would all come at the full
935 minute and disturb each other. Besides that I'd hate to have a backbone
936 with more than a few seconds latency.
938 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
940 * a second rsync server with access control for PAUSE. Port? 873 is the
941 standard port, let's take 8873.
943 * if there were a filesystem based on this, it would have a slow access
944 to inexistent files. It would probably provide wrong readdir (only based
945 on current content) or also a slow one (based on a recentfile written
946 after the call). But it would provide fast access to existing files. Or
947 one would deliberately allow slightly blurred answers based on some
948 sqlite reflection of the recentfiles.
950 * todo: write a variant of mirror() that combines two or more
951 recentfiles and treats them like one
953 * todo: signal handler to remove the tempfile
955 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
957 * now that we have the symlink I forgot how it should be used in
960 * the z loop: add missing files to Z file. Just append them (instead of
961 prepending). So one guy prepends something from the Y file from time to
962 time and another guy appends something rather frequently. Collecting
963 pond. When Y merges into Z, things get epoch and the collecting pond
964 gets smaller. What exactly are "missing files"?
966 take note of current epoch of the alpha file, let's call it the
969 find all files on disk
971 remove all files registered in the recentworld up to recent-ts
973 remove all files that have been deleted after recent-ts according to
976 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
978 * rersyncrecent might be a cronjob with a (locked) state file which
979 contains things like after and maybe last z sync or such?
981 rrr-mirror might be an alternative name but how would we justify the
982 three Rs when there is no Re-Rsync-Recent?
984 With the --loop parameter it is an endless loop, without it is no loop.
985 At least this is simple.
987 * todo: new accssor z-interval specifies how often the Z file is updated
988 against the filesystem. We probably want no epoch stamp on these
989 entries. And we want to be able to filter the entries (e.g. no
990 by-modules and by-category tree)
992 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
994 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
995 into Z. We do not need the other files filled up with filesystem
998 * need interface to query for a file in order to NOT call update on
999 PAUSE a second time within a short time.
1001 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1003 * recommended update interval? Makes no sense, is different for
1010 change-log-default-name: "Todo"