1 2009-03-17 Andreas J. Koenig <andk@cpan.org>
3 * done: verified the existence of the floating point bug in bleadperl
4 and verified that switching from YAML::Syck to YAML::XS does not resolve
7 BTW, the switch was doable with
9 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
11 and should be considered as a separate TODO
13 * todo: integrate a dirty update with two aggregate calls before
14 unlocking for frictionless dirtying
16 * todo: start the second rsync daemon on pause
18 * todo: move index files to .recent: this cannot simply be done by
19 setting filenameroot to .recent/RECENT. Other parts of the modules rely
20 on the fact that dirname(recentfile) is the root of the mirrored tree.
22 2009-03-16 Andreas J. Koenig <andk@cpan.org>
24 * What was the resolution of the mirror.pl delete hook bug? Do we call
25 the delete hook when pause removes a file from MUIR?
27 * Today on pause: Updating 2a13fba..29f284d and installing it for
28 /usr/local/perl-5.10.0{,-RC2}
30 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
31 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
33 2009-03-15 Andreas J. Koenig <andk@cpan.org>
35 * currently recent_events has the side effect of setting dirtymark
36 because it forces a read on the file. That should be transparent, so
37 that the dirtymark call always forces a cache-able(?) read.
39 * The bug below is -- after a lot of trying -- not reproducible on a
40 small script, only in the large test script. The closest to the output
46 my $x = "01237123229.8814";
51 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
53 $l = "1237123231.22458";
54 $r = "1237123231.22458";
58 Devel::Peek::Dump $x = $l <=> $r;
62 The checked in state at c404a85 fails the test with my
63 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
64 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
65 in the test script always consistent.
67 * Todo: write a test that inserts a second dirty file with an already
68 existing timestamp. DONE
70 * Bug in perl 5.10 on my 64bit box:
72 DB<98> Devel::Peek::Dump $l
73 SV = PVMG(0x19e0450) at 0x142a550
75 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
78 PV = 0x194ce70 "1237123231.22458"\0
82 DB<99> Devel::Peek::Dump $r
83 SV = PVMG(0x19e0240) at 0x142a3e8
85 FLAGS = (PADMY,POK,pPOK)
88 PV = 0x19ff900 "1237123231.22458"\0
92 DB<100> Devel::Peek::Dump $l <=> $r
93 SV = IV(0x19ea6e8) at 0x19ea6f0
95 FLAGS = (PADTMP,IOK,pIOK)
98 DB<101> Devel::Peek::Dump $l
99 SV = PVMG(0x19e0450) at 0x142a550
101 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
103 NV = 1237123231.22458
104 PV = 0x194ce70 "1237123231.22458"\0
108 DB<102> Devel::Peek::Dump $r
109 SV = PVMG(0x19e0240) at 0x142a3e8
111 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
113 NV = 1237123231.22458
114 PV = 0x19ff900 "1237123231.22458"\0
118 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
120 * Todo: reset "done" or "covered" and "minmax" after a dirty operation?
122 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
124 * $obj->merge ($other) needs to learn about equal epoch which may happen
125 since dirty_epoch intruded.
127 * Wontfix anytime soon: I think we currently do not support mkdir. Only
130 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
132 * Todo: continue working on update(...,$dirty_epoch). It must be
133 followed by a fast_aggregate!
135 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
137 * maybe we need a closest_entry or fitting_interval or something like
138 that. We want to merge an event into the middle of some recentfile.
139 First we do not know which file, then we do not know where to lock,
140 where to enter the new item, when and where to correct the dirtymark.
142 So my thought is we should first find which file.
144 Another part of my brain answers: what would happen if we would enter
145 the new file into the smallest file just like an ordinary new event,
146 just as an old event?
148 (1) we would write a duplicate timestamp? No, this would be easy to
151 (2) we would make the file large quickly? Yes, but so what? We are
152 changing the dirtymark, so are willing to disturb the downstream hosts.
154 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
156 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
160 https://rt.cpan.org/Ticket/Display.html?id=41199
162 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
164 Strangest output being HAKANARDO who managed to upload
166 Here is a better oneliner that includes also the first line of each
169 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$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)}'
171 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
173 * memory leak in the syncher? It currently weighs 100M.
177 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
180 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
182 * lookup by epoch and by path and use this ability on the pause to never
183 again register a file twice that doesn't need it. Let's call it
186 * after the dirtymark is done: fill up recentfiles with fake (historic)
187 entries; fill up with individual corrections; algorithm maybe to be done
188 with bigfloat so that we can always place something in the middle
189 between two entries. Before we must switch to bigfloat we could try to
190 use Data::Float::nextup to get the.
192 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
195 * dirtymark now lives in Recentfile, needs to be used in rmirror.
197 * find out why the downloader died after a couple of hours without a net
198 connection. Write a test that survives the not-existence of the other
201 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
203 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
204 it makes less sense with tight coupling for secondary files. But asking
205 the server what the current epoch is might be cheaper on HTTP than on
206 rsync. (Needs to be evaluated)
208 * remove the 0.00 from the verbose overview in the Merged column in the
211 * write tests that expose the problems of the last few days: cascading
212 client/server roles, tight coupling for secondary RFs, deletes after
215 * Some day we might want to have policy options for the slave:
216 tight/loose/no coupling with upstream for secondary RFs. tight is what
217 we have now. loose would wait until a gap occurs that can be closed.
219 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
221 * revisit all $rfs->[$i+1] places if they now make sense still
223 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
225 * another bug is the fact that the mirror command deletes files before
226 it unhides the index file, thus confusing downstream slaves. We must not
227 delete before unhiding and must delete after unhiding. FIXED.
229 * new complication about the slave that is playing a server role.
230 Currently we mirror from newest to oldest with a hidden temporary file
231 as index. And when one file is finished, we unhide the index file.
232 Imagine the cascading server/slave is dead for a day. It then starts
233 mirroring again with the freshest thing and unhides the freshest index
234 file when it has worked through it. In that moment it exposes a time
235 hole. Because it now works on the second recentfile which is still
238 We currently do nothing special to converge after such a drop out. At
239 least not intentionally and robustly and thought through.
241 The algorithm we use to seed the next file needs quite a lot of more
242 robustness than it currently has. Something to do with looking at the
243 merged element of the next rf and when it has dropped off, we seed
244 immediately. And if it ramains dropped off, we seed again, of course.
246 Nope, looking from smaller to larger RFS we look at the merged element
247 of this RF and at the minmax/max element of the next RF. If that
248 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
251 And we need a public accessor seed and unseed or seeded. But not the mix
252 of public and private stuff that then is used behind the back.
254 And then the secondary* stuff must go.
256 And we must understand what the impact is on the DONE system. Can it go
257 unnoticed that there was a hole? And could the DONE system have decided
258 the hole is covered? This should be testable with three directories where
259 the middle stops working for a while. Done->merge is suspicious, we must
260 stop it from merging non-conflatable neighbors due to broken continuity.
264 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
266 * Slaven suggests to have the current epoch or the whole current
267 recentfile available from the HTTP server and take it away with
268 keepalive. This direction goes the granularity down to subseconds.
270 We might want to rewrite everything to factor out transport and allow
271 the whole thing to run via HTTP.
273 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
275 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
278 * smoker on k81 fetching from k75 to verify cascading works. See
279 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
281 * maybe the loop should wait for CHECKSUMS file after every upload. And
282 CPAN.pm needs to deal with timestamps in the future.
284 * do not forget the dirtymark!
286 Text: have a new flag on recentfiles with the meaning: if this
287 changes, you're required to run a full rsync over all the files. The
288 reason why we set it would probably be: some foul happened. we injected
289 files in arbitrary places or didn't inject them although they changed.
290 The content of the flag? Timestamp? The relation between the
291 recentfiles would have to be inheritance from the principal, because any
292 out of band changes would soon later propagate to the next recentfile.
294 By upping the flag often one can easily ruin the slaves.
296 last out of band change? dirtymark?
298 Anyway, this implies that we read a potentially existing recentfile
301 And it implies that we have an eventloop that keeps us busy in 2-3
302 cycles, one for current stuff (tight loop) and one for the recentfiles
303 (cascade when principal has changed), one for the old stuff after a
306 And it implies that the out-of-band change in any of the recentfiles
307 must have a lock on the principal file and there is the place to set the
310 * start a FAQ, especially quick start guide questions. Also to aid those
311 problematic areas where we have no good solution, like the "links"
314 * wish feedback when we are slow.
318 * Remove a few DEBUG statements.
320 * The multiple-rrr way of doing things needs a new option to rmirror,
321 like piecemeal or so. Not urgent because after the first pass through,
322 things run smoothely. It's only ugly during the first pass.
324 * I have the suspicion that the code is broken that decides if the
325 neighboring RF needs to be seeded. I fear when too much time has gone
326 between two calls (in our case more than one hour), it would not seed
327 the neighbor. Of course this will never be noticed, so we need a good
330 * local/localroot confusion: I currently pass both options but one must
333 * accounts for early birds on PAUSE rsync daemon.
335 * hardcoded 20 seconds
337 * who mirrors the index? DOING now.
339 * which CPAN mirrors offer rsync?
341 * visit all XXX, visit all _float places
343 * rename the pathdb stuff, it's too confusing. No idea how.
345 * rrr-inotify, backpan, rrr-register
347 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
349 * current bugs: the pathdb seems to get no reset, the seeding of the
350 secondaryttl stuff seems not to have an effect. Have helped myself with
351 a rand(10), need to fix this back. So not checked in. Does the rand
354 The rand thing helps. The secondaryttl stuff was in the wrong line,
357 The pathdb stuff was because I called either _pathdb or __pathdb on the
358 wrong object. FIXED now.
360 * It's not so beautiful if we never fetch the recentfiles that are not
361 the principal, even if this is correct behaviour. We really do not need
362 them after we have fetched the whole content.
364 OK, we want a switch for that: secondaryttl DONE
366 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
368 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
369 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
370 it lets 40 follow. FIXED
372 * See that the long running process really only updates the principal
373 file unless it has missed a timespan during which something happened. If
374 nothing happened, it must notice even when it misses the timespan. DONE
376 * we must throw away the pathdb when we have reached the end of Z. From
377 that moment we can have a very small pathdb because the only reason for
378 a pathdb is that we know to ignore old records in old files. We won't
379 need this pathdb again before the next full pass over the data is
380 necessary and then we will rebuild it as we go along. DONE
382 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
384 * I think, Done::register_one is doing wrong in that it does not
385 conflate neighboring pieces. The covered() method cannot do this because
386 it has no recent_events array at hand. But register_one has it and could
387 do it and for some reason misses to do it (sometimes).
389 This means that the three tests I just wrote can probably not survive
390 because they test with an already broken Done structure.
392 The art now is to detect how it happens, then to reproduce, then write a
395 So from the logfile this is what happens: we have a good interval with
396 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
397 top of it. Locally we now mirror F2 and open a new done interval for it.
398 Then we mirror F1 but this time with the timestamp T1b. And when we then
399 try to close the gap, we do not find T1 but instead something older. We
400 should gladly accept this older piece and this would fix this bug.
404 * bug to fix: when the 1h file changes while rmirror is running, we do
405 correctly sync the new files but never switch to the 6h file but rather
406 stay in a rather quick loop that fetches the 1h file again and again.
408 Is it possible that we initialize a new object? Or does
409 get_remote_recentfile_as_tempfile overwrite something in myself?
411 Want a new option: _runstatusfile => $file which frequently dumps the
412 state of all recentfiles to a file.
416 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
418 * Todo: now teach update to verify the timestamp is about to write
419 against the previous and use _increase_a_bit if it doesn't comply with
420 strict monotony. DONE
422 * The problem of rounding. So far perl's default precision was
423 sufficient. One day it won't be. FakeFloat has an easy job when it is
424 only reading and other machines have written correctly. But when we want
425 to write a floating point number that is a bit larger than the other
426 one, then we need our own idea of precision.
428 Slaven said: just append a "1". This might be going towards the end of
429 usability too quickly. I'd like something that actually uses the decimal
430 system. Well, appending a 1 also does this but...
432 E.g. we have 1.0. nextup on this architecture is starting with
433 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
434 taken the 1.0000000000000003 and the next user comes and the time tells
435 him 1.0 again. He has to beat my number without stepping over the
436 nextup. This is much less space than I had when I chose 1,2,3.
438 What is also irritating is that nextup is architecture dependent. The
439 128 bit guy must choose very long numbers to fit in between whereas the
440 other one with 16 bit uses larger steps. But then the algorithm is the
441 same for both, so that would be a nice thing.
443 I see two situation where we need this. One is when Time::HiRes returns
444 us a value that is <= the last entry in our recentfile. In this case
445 (let's call it the end-case) we must fill the region between that number
446 and the next higher native floating point number. The other is when we
447 inject an old file into an old recentfile (we would then also set a new
448 dirtymark). We find the integer value already taken and need a slightly
449 different one (let's call it the middle-case). The difference between
450 the two situations is that the next user will want to find something
451 higher than my number in the end-case and something lower than my number
454 So I suggest we give the function both a value and an upper bound and it
455 calculates us a primitive middle. The upper bound in the middle-case is
456 the next integer. The upper bound on the end-case is the nextup floating
457 point number. But the latter poses another problem: if we have occupied
458 the middle m between x and nextup(x), then the nextup(m) will probably
459 not be the same as nextup(x) because some rounding will take place
460 before the nextup is calculated and when the rounding reaches the
461 nextup(x), we will end up at nextup(nextup(x)).
463 So we really need to consider the nextup and the nextdown from there and
464 then the middle and that's the number we may approach asymptotically.
467 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
469 * consider deprecating the use of RECENT.recent as a symlink. It turns
470 out to need extra hoops with the rsync options and just isn't worth it.
471 Or maybe these extra hoops are needed anyway for the rest of the tree?
472 Nope, can't be the case because not all filesystems support symlinks.
474 But before doing the large step, I'll deprecate the call of
475 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
476 only introduced to resolve RECENT.recent and complicates the routine far
477 beyond what it deserves.
479 DONE. Won't deprecate RECENT.recent, just moved its handling to the
482 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
484 * I think it's a bug that the rsync_option links must be set to true in
485 order to support RECENT.recent and that nobody cares to set it
486 automatically. Similar for ignore_link_stat_errors.
488 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
490 * Todo: find all todos together and make a plan what is missing for a
493 - verifytree or something like that. fsck maybe.
495 - rersyncrecent, the script itself? What it do?
497 - a way to only mirror the recentfiles without mirroring the whole
498 remote system such that people can decide to mirror only partially see
499 also 2008-08-30. .shadow-xxx directory? this also needed for a
500 filesystem that is still incomplete and might need the mirrorfiles for
503 - long living objects that mirror again and again. Inject something
504 into ta, see how it goes over to tb.
506 - how do we continue filling up the DONE system when we use an object
507 for the second time? "fully covered" and "uptodate" or new terminology.
509 - overview called on the wrong file should be understandable
511 - the meta data field that must change when we fake something up so that
512 the downstream people know they have to re-fetch everything.
514 - how tolerant are we against missing files upstream? how do we keep
515 track? there are legitimate cases where we did read upstream index right
516 before a file got deleted there and then find that file as new and want
517 it. There are other cases that are not self healing and must be tracked
520 - how, exactly, do we have to deal with deletes? With rsync errors?
522 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
523 authors) failed: No such file or directory (2)
525 The file above is a delete in 1h and a new in file 1M and the
526 delete in the locally running rmirror did not get propagated to the 1M
527 object. Bug. And the consequence is a standstill.
529 It seems that a slave that works with a file below the principal needs
530 to merge things all the way up to get rid of later deletes. Or keep
531 track of all deletes and skip them later. So we need a trackdeletes.pm
532 similar to the done.pm?
534 see also 2008-08-20 about spurious deletes that really have no add
535 counterpart and yet they are not wrong.
537 - consider the effect when resyncing the recentfile takes longer than
538 the time per loop. Then we never rsync any file. We need to diagnose
539 that and force an increase of that loop time. But when we later are fast
540 enough again because the net has recovered, then we need to switch back
541 to original parameters. ERm, no, it's enough to keep syncing at least
542 one file before refetching an index file.
544 - remember to verify that no temp files are left lying around and the
547 - status file for not long running jobs that want to track upstream with
550 - revisit all XXX _float areas and study Sub::Exporter DONE
552 - persistent DB even though we just said we do not need it. Just for
553 extended capabilities and time savings when, for example, upstream
554 announces a reset and we get new recentfiles and could then limit
555 ourselves to a subset of files (those that have a changed epoch) in a
556 first pass and would only then do the loop to verify the rest. Or
559 * Todo: aggregate files should know their feed and finding the principal
560 should be done stepwise. (?)
562 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
563 left locks around. DONE
565 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
567 * maybe extend the _overview so that it always says if and where the
568 last file is in the next file and where the next event in the next rf
569 would lie. No, don't like this anymore. REJECT
571 * take the two new redundant tests out again, only the third must
574 * Todo: add a sanity check if the merged structure is really pointing to
575 a different rf and that this different rf is larger. DONE
577 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
579 * now test, if they are overlapping. And test if there is a file in the
580 next rf that would fit into this rf's interval.
582 1h 1222324012.8474 1222322541.7963 0.4086
583 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
584 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
585 1W 1222313218.3626 1221708477.5829 167.9835
587 I suspect that somebody writes a merged timestamp without having merged
588 and then somebody else relies on it.
590 If aggregate is running, the intervals must not be extravagated, if it
591 is not running, there must not be bounds, the total number of events in
592 the system must be counted and must be controlled throughout the tests.
593 That the test required the additional update was probably nonsense,
594 because aggregate can cut pieces too. FIXED & DONE
596 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
598 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
601 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
603 * Most apparent bug at the moment is that the recentfiles are fetched
604 too often. Only the principal should be fetched and if it has not
605 changed, the others should not be refetched. ATM I must admit that I'm
606 happy that we refetch more often than needed because I can more easily
607 fix bugs while the thing is running.
609 * Let's say, 1220474966.19501 is a timestamp of a file that is already
610 done but the done system does not know about it. The reason for the
611 failure is not known and we never reach the status uptodate because of
612 this. We must get over it.
614 Later it turns out that the origin server had a bug somewhere.
615 1220474966.19042 came after 1220474966.19501. Or better: it was in the
616 array of the recentfile one position above. The bug was my own.
618 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
620 * There is the race condition where the server does a delete and the
621 slave does not yet know and then tries to download it because he sees
622 the new. So for this time window we must be more tolerant against
623 failure. If we cannot download a file, we should just skip it and should
624 not retry immediately. The whole system should discover the lost thing
625 later. Keeping track with the DONE system should really be a no brainer.
627 But there is something more: the whole filesystem is a database and the
628 recentfiles are one possible representation of it. It's a pretty useful
629 representation I think that's why I have implemented something around
630 it. But for strictly local operation it has little value. For local
631 operation we would much rather have a database. So we would enter every
632 recentfile reading and every rsync operation and for every file the last
633 state change and what it leads to. Then we would always ignore older
634 records without the efforts involved with recentfiles.
636 The database would have: path,recentepoch,rsyncedon,deletedon
638 Oh well, not yet clear where this leads to.
640 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
642 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
643 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
645 * statusfile, maybe only on demand, alone to have a sharp debugging
646 tool. It is locked and all recentfiles dump themselves into it and we
647 can build a viewer that lets us know where we stand and what's inside.
649 * remember: only the principal recentfile needs expiration, all others
650 shall be expired by principal if it discovers that something has move
653 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
655 * Always check if we stringify to a higher value than in the entry
658 * And in covered make an additional check if we would be able to see a
659 numerical difference between the two numbers and if we can't then switch
660 to a different, more expensive algorithm. Do not want to be caught by
661 floating surprises. DONE
663 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
665 * caching has several aspects here: we can cache the interval of the
666 recentfile which only will change when the mtime of the file changes. We
667 must re-mirror the recentfile when its ttl has expired. Does have_read
668 tell you anything? It counts nothing at all. Only the mtime is
669 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
670 thing the have_mirrored because it is unrelated to the mtime.
672 * Robustness of floating point calculations! I always thought that the
673 string calculated by the origin server for the floating representation
674 of the epoch time is just a string. When we convert it to a number and
675 later back to a string, the other computer might come to a different
676 conclusion. This must not happen, we want to preserve it under any
677 circumstances. I will have to write tests with overlong sequences that
678 get lost in arithmetic and must see if all still works well. DONE
680 But one fragile point remains: if one host considers a>b and the other
681 one considers them == but no eq. To prevent this, we must probably do
682 some extra homework. DONE
684 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
686 * the concept of tracking DONE needs an object per recentfile that has
687 something like these methods:
689 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
691 covered() register() covered()
693 The unclear thing is how we translate points in time into intervals. We
694 could pass a reference to the current recent_events array when running
695 we_have(xxx) and let the DONE object iterate over it such that it only
696 has to store a list of intervals that can melt into each other. Ah, even
697 passing the list together with a list of indexes seems feasiable.
699 Or maybe ask for the inverted list?
701 Whenever the complete array is covered by the interval we say we are
702 fully covered and if the recentfile is not expired, we are uptodate.
704 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
706 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
708 * need a way to "return" the next entry after the end of a list. When
709 the caller says "before" or "after" we would like to know if he could
710 cover that interval/threshold or not because this influences the effect
711 of a newer timestamp of that recentfile. DONE with $opt{info}.
713 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
715 * one of the next things to tackle: the equivalent of csync2 -TIXU.
717 loop implies tixu (?). Nope, something like --statefile decides. Per
720 T test, I init, X including removals, U nodirtymark
722 So we have no concept of dirtymarks, we only trust that since we are
723 running we have observed everything steadily. But people will not let
724 this program run forever so we must consider both startup penalty and
725 book keeping for later runs. We keep this for later. For now we write a
726 long running mirror that merges several intervals.
728 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
730 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
733 * rersyncrecent, the script: default to one week. The name of the switch
734 is --after. Other switches? --loop!
736 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
738 * need a switch --skip-deletes (?)
740 * need a switch --enduser that tells us that the whole tempfile
741 discipline is not needed when there is no downstream user. (?)
743 Without this switch we cannot have a reasonable recent.pl that just
744 displays the recent additions. Either we accept to download everything.
745 Or we download temporary files without the typical rsync protocol
748 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
749 File::Temp::tempdir, this might be a win.
751 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
753 * apropos missing: we have no push, we never know the downstream
754 servers. People who know their downstream hosts and want to ascertain
755 something will want additional methods we have never thought about, like
756 update or delete a certain file.
758 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
760 * tempted to refactor rmirror into resolve_symlink, localize, etc.
761 Curious if rsync_options=links equal 0 vs. 1 will make the expected
764 * rsync options: it's a bit of a pain that we usually need several rsync
765 options, like compress, links, times, checksum and that there is no
766 reasonable default except the original rsync default. I think wee can
767 safely assume that the rsync options are shared between all recentfile
768 instances within one recent tree.
770 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
772 * deletes: if a delete follows an add quickly enough it may happen that
773 a downstream mirror did not see the add at all! It seems this needs to
774 be mentioned somewhere. The point here is that even if the downstream is
775 never missing the principal timeframe it may encounter a "delete" that
776 has no complimentary "add" anywhere.
778 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
780 * I suspect the treat of metadata is incorrect during read or something.
781 The bug that I am watching is that between 06:08 and 06:09 the 6h file
782 contained more than 6 hours worth of data. At 06:08 we merged into the
783 1d file. We need to take snapshots of the 6h file over the course of an
784 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
787 Much worse: watching the 1h file: right at the moment (at 06:35) it
788 covers 1218867584-1219120397 which is 70 hours.
790 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
791 08:19:44 2008, that is when I checked out last time, so it seems to be
792 aggregating and never truncating?
794 No, correct is: it is never truncating; but wrong is: it is aggregating.
795 It does receive a lot of events from time to time from a larger file.
796 Somehow a large file gets merged into the small one and because the
797 "meta/merged" attribute is missing, nobody is paying attention. I
798 believe that I can fix this by making sure that metadata are honoured
799 during read. DONE and test adjusted.
801 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
803 * grand renaming plan
805 remotebase => remoteroot to fit well with localroot DONE
806 local_path() => localroot seems to me should already work DONE
807 recentfile_basename => rfilename no need to stress it has no slash DONE
809 filenameroot??? Doesn't seem too bad to me today. Maybe something like
810 kern? It would anyway need a deprecation cycle because it is an
811 important constructor.
813 * I like the portability that Data::Serializer brings us but the price
814 is that some day we might find out that it is slowing us a bit. We'll
817 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
819 * should we not enter the interval of the principal (or the interval of
820 the merging file?) in every aggregated/merged file?
822 * we should aim at a first release and give up on thinking about
823 sanitizing stuff and zloop. Let's just admit that a full traditional
824 rsync is the only available sanitizer ATM. Otherwise it's complicated
825 stuff: sanitizing on the origin server, sanitizing on the slaves,
826 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
827 get the basics out before this becomes a major cause for mess.
829 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
831 * On OSes not supporting symlinks we expect that RECENT.recent contains
832 the contents of the principal recentfile. Actually this is identical on
833 systems supporting symlinks. Simple, what follows from that is that we
834 need to keep the serializer in the metadata because we cannot read it
835 from the filename, doesn't it? Of course not. It's a chicken and egg
836 problem. This leaves us with the problem to actually parse the
837 serialized data to find out in which format it is. So who can do the 4
838 or 5 magics we wanted to support? File::LibMagic?
840 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
842 * remotebase and recentfile_basename are ugly names. Now that we need a
843 word for the shortest/principal/driving recentfile too we should do
846 localroot is good. rfile is good. local_path() is bad, local_path($path)
847 is medium, filenameroot() is bad, remotebase is bad, recentfile is
850 Up to now remotebase was the string that described the remote root
851 directory in rsync notation, like pause.perl.org::authors. And
852 recentfile_basename was "RECENT-1h.yaml".
854 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
856 * The test that was added in today's checkin is a good start for a test
857 of rmirror. We should have more methods in Recent.pm: verify,
858 addmissingfiles. We should verify the current tree, then rmirror it and
859 then verifytree the copy. We could then add some arbitrary file and let
860 it be discovered by addmissingfiles, then rmirror again and then
861 verifytree the copy again.
863 Then we could start stealing from csync2 sqlite database [no port to
864 OSX!] and fill a local DB. And methods to compare the database with the
865 recentfiles. Our strength is that in principle we could maintain state
866 with a single float. We have synced up to 1234567890.123456. If the Z
867 file does not add new files all we have to do is mirror the new ones and
870 This makes it clear that we should extend current protocol and declare
871 that we cheat when we add files too late, just to help the other end
872 keeping track. Ah yes, that's what was meant when zloop was mentioned
875 Maybe need to revisit File::Mirror to help me with this task.
877 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
879 * There must be an allow-me-to-truncate flag in every recentfile.
880 Without it one could construct a sequence of updates winning the locking
881 battle against the aggregator. Only if an aggregator has managed to
882 merge data over to the next level, truncating can be allowed. DONE with
885 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
887 * We should probably guarantee that no duplicates enter the aggregator
890 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
892 * To get merge operation faster would need a good benchmark test. What
893 02 spits out isn't reliable enough and is dominated by many other
896 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
897 Author: Andreas J. Koenig <k@k75.(none)>
898 Date: Sat Aug 2 07:58:04 2008 +0200
902 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
903 Author: Andreas J. Koenig <k@k75.(none)>
904 Date: Sat Aug 2 11:40:29 2008 +0200
906 there was a lot of trying but the effect is hardly measurable with
909 * overhead of connecting seems high. When setting
910 max_files_per_connection to 1 we see that.
912 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
914 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
916 25138.2215600014/3600 = 6.98283932222261
918 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
919 be a bug in the aggregator. FIXED
921 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
923 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
924 we have it, can we mirror it, mirror it!
926 I fear this needs a new class which might be called
927 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
928 and treat them as an entity. I realize that a single recentfile may be
929 sufficient for certain tasks and that it is handy for the low level
930 programmer but it is not nice to use. If there is a delete in the 1h
931 file then the 6h file still contains it. Seekers of the best information
932 need to combine at least some of the recentfiles most of the time.
934 There is the place for the Z loop!
936 But the combination is something to collect in a database, isn't it. Did
937 csync2 just harrumph?
939 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
941 * it just occurred to me that hosts in the same mirroring pool could
942 help out each other even without rewriting the recentfile. Just fetch
943 the stuff to mirror from several places, bingo. But that's something
944 that should rather live in a separate package or in rsync directly.
946 * cronjobs are unsuited because with ntp they would all come at the full
947 minute and disturb each other. Besides that I'd hate to have a backbone
948 with more than a few seconds latency.
950 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
952 * a second rsync server with access control for PAUSE. Port? 873 is the
953 standard port, let's take 8873.
955 * if there were a filesystem based on this, it would have a slow access
956 to inexistent files. It would probably provide wrong readdir (only based
957 on current content) or also a slow one (based on a recentfile written
958 after the call). But it would provide fast access to existing files. Or
959 one would deliberately allow slightly blurred answers based on some
960 sqlite reflection of the recentfiles.
962 * todo: write a variant of mirror() that combines two or more
963 recentfiles and treats them like one
965 * todo: signal handler to remove the tempfile
967 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
969 * now that we have the symlink I forgot how it should be used in
972 * the z loop: add missing files to Z file. Just append them (instead of
973 prepending). So one guy prepends something from the Y file from time to
974 time and another guy appends something rather frequently. Collecting
975 pond. When Y merges into Z, things get epoch and the collecting pond
976 gets smaller. What exactly are "missing files"?
978 take note of current epoch of the alpha file, let's call it the
981 find all files on disk
983 remove all files registered in the recentworld up to recent-ts
985 remove all files that have been deleted after recent-ts according to
988 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
990 * rersyncrecent might be a cronjob with a (locked) state file which
991 contains things like after and maybe last z sync or such?
993 rrr-mirror might be an alternative name but how would we justify the
994 three Rs when there is no Re-Rsync-Recent?
996 With the --loop parameter it is an endless loop, without it is no loop.
997 At least this is simple.
999 * todo: new accssor z-interval specifies how often the Z file is updated
1000 against the filesystem. We probably want no epoch stamp on these
1001 entries. And we want to be able to filter the entries (e.g. no
1002 by-modules and by-category tree)
1004 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1006 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1007 into Z. We do not need the other files filled up with filesystem
1010 * need interface to query for a file in order to NOT call update on
1011 PAUSE a second time within a short time.
1013 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1015 * recommended update interval? Makes no sense, is different for
1022 change-log-default-name: "Todo"