1 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * maybe we need a closest_entry or fitting_interval or something like
4 that. We want to merge an event into the middle of some recentfile.
5 First we do not know which file, then we do not know where to lock,
6 where to enter the new item, when and where to correct the dirtymark.
8 So my thought is we should first find which file.
10 Another part of my brain answers: what would happen if we would enter
11 the new file into the smallest file just like an ordinary new event,
14 (1) we would write a duplicate timestamp? No, this would be easy to
17 (2) we would make the file large quickly? Yes, but so what? We are
18 changing the dirtymark, so are willing to disturb the downstream hosts.
20 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
22 * 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
26 https://rt.cpan.org/Ticket/Display.html?id=41199
28 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
30 Strangest output being HAKANARDO who managed to upload
32 Here is a better oneliner that includes also the first line of each
35 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)}'
37 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
39 * memory leak in the syncher? It currently weighs 100M.
43 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
46 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
48 * lookup by epoch and by path and use this ability on the pause to never
49 again register a file twice that doesn't need it. Let's call it
52 * after the dirtymark is done: fill up recentfiles with fake (historic)
53 entries; fill up with individual corrections; algorithm maybe to be done
54 with bigfloat so that we can always place something in the middle
55 between two entries. Before we must switch to bigfloat we could try to
56 use Data::Float::nextup to get the.
58 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
61 * dirtymark now lives in Recentfile, needs to be used in rmirror.
63 * find out why the downloader died after a couple of hours without a net
64 connection. Write a test that survives the not-existence of the other
67 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
69 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
70 it makes less sense with tight coupling for secondary files. But asking
71 the server what the current epoch is might be cheaper on HTTP than on
72 rsync. (Needs to be evaluated)
74 * remove the 0.00 from the verbose overview in the Merged column in the
77 * write tests that expose the problems of the last few days: cascading
78 client/server roles, tight coupling for secondary RFs, deletes after
81 * Some day we might want to have policy options for the slave:
82 tight/loose/no coupling with upstream for secondary RFs. tight is what
83 we have now. loose would wait until a gap occurs that can be closed.
85 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
87 * revisit all $rfs->[$i+1] places if they now make sense still
89 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
91 * another bug is the fact that the mirror command deletes files before
92 it unhides the index file, thus confusing downstream slaves. We must not
93 delete before unhiding and must delete after unhiding. FIXED.
95 * new complication about the slave that is playing a server role.
96 Currently we mirror from newest to oldest with a hidden temporary file
97 as index. And when one file is finished, we unhide the index file.
98 Imagine the cascading server/slave is dead for a day. It then starts
99 mirroring again with the freshest thing and unhides the freshest index
100 file when it has worked through it. In that moment it exposes a time
101 hole. Because it now works on the second recentfile which is still
104 We currently do nothing special to converge after such a drop out. At
105 least not intentionally and robustly and thought through.
107 The algorithm we use to seed the next file needs quite a lot of more
108 robustness than it currently has. Something to do with looking at the
109 merged element of the next rf and when it has dropped off, we seed
110 immediately. And if it ramains dropped off, we seed again, of course.
112 Nope, looking from smaller to larger RFS we look at the merged element
113 of this RF and at the minmax/max element of the next RF. If that
114 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
117 And we need a public accessor seed and unseed or seeded. But not the mix
118 of public and private stuff that then is used behind the back.
120 And then the secondary* stuff must go.
122 And we must understand what the impact is on the DONE system. Can it go
123 unnoticed that there was a hole? And could the DONE system have decided
124 the hole is covered? This should be testable with three directories where
125 the middle stops working for a while. Done->merge is suspicious, we must
126 stop it from merging non-conflatable neighbors due to broken continuity.
130 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
132 * Slaven suggests to have the current epoch or the whole current
133 recentfile available from the HTTP server and take it away with
134 keepalive. This direction goes the granularity down to subseconds.
136 We might want to rewrite everything to factor out transport and allow
137 the whole thing to run via HTTP.
139 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
141 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
144 * smoker on k81 fetching from k75 to verify cascading works. See
145 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
147 * maybe the loop should wait for CHECKSUMS file after every upload. And
148 CPAN.pm needs to deal with timestamps in the future.
150 * do not forget the dirtymark!
152 Text: have a new flag on recentfiles with the meaning: if this
153 changes, you're required to run a full rsync over all the files. The
154 reason why we set it would probably be: some foul happened. we injected
155 files in arbitrary places or didn't inject them although they changed.
156 The content of the flag? Timestamp? The relation between the
157 recentfiles would have to be inheritance from the principal, because any
158 out of band changes would soon later propagate to the next recentfile.
160 By upping the flag often one can easily ruin the slaves.
162 last out of band change? dirtymark?
164 Anyway, this implies that we read a potentially existing recentfile
167 And it implies that we have an eventloop that keeps us busy in 2-3
168 cycles, one for current stuff (tight loop) and one for the recentfiles
169 (cascade when principal has changed), one for the old stuff after a
172 And it implies that the out-of-band change in any of the recentfiles
173 must have a lock on the principal file and there is the place to set the
176 * start a FAQ, especially quick start guide questions. Also to aid those
177 problematic areas where we have no good solution, like the "links"
180 * wish feedback when we are slow.
184 * Remove a few DEBUG statements.
186 * The multiple-rrr way of doing things needs a new option to rmirror,
187 like piecemeal or so. Not urgent because after the first pass through,
188 things run smoothely. It's only ugly during the first pass.
190 * I have the suspicion that the code is broken that decides if the
191 neighboring RF needs to be seeded. I fear when too much time has gone
192 between two calls (in our case more than one hour), it would not seed
193 the neighbor. Of course this will never be noticed, so we need a good
196 * local/localroot confusion: I currently pass both options but one must
199 * accounts for early birds on PAUSE rsync daemon.
201 * hardcoded 20 seconds
203 * who mirrors the index? DOING now.
205 * which CPAN mirrors offer rsync?
207 * visit all XXX, visit all _float places
209 * rename the pathdb stuff, it's too confusing. No idea how.
211 * rrr-inotify, backpan, rrr-register
213 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
215 * current bugs: the pathdb seems to get no reset, the seeding of the
216 secondaryttl stuff seems not to have an effect. Have helped myself with
217 a rand(10), need to fix this back. So not checked in. Does the rand
220 The rand thing helps. The secondaryttl stuff was in the wrong line,
223 The pathdb stuff was because I called either _pathdb or __pathdb on the
224 wrong object. FIXED now.
226 * It's not so beautiful if we never fetch the recentfiles that are not
227 the principal, even if this is correct behaviour. We really do not need
228 them after we have fetched the whole content.
230 OK, we want a switch for that: secondaryttl DONE
232 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
234 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
235 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
236 it lets 40 follow. FIXED
238 * See that the long running process really only updates the principal
239 file unless it has missed a timespan during which something happened. If
240 nothing happened, it must notice even when it misses the timespan. DONE
242 * we must throw away the pathdb when we have reached the end of Z. From
243 that moment we can have a very small pathdb because the only reason for
244 a pathdb is that we know to ignore old records in old files. We won't
245 need this pathdb again before the next full pass over the data is
246 necessary and then we will rebuild it as we go along. DONE
248 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
250 * I think, Done::register_one is doing wrong in that it does not
251 conflate neighboring pieces. The covered() method cannot do this because
252 it has no recent_events array at hand. But register_one has it and could
253 do it and for some reason misses to do it (sometimes).
255 This means that the three tests I just wrote can probably not survive
256 because they test with an already broken Done structure.
258 The art now is to detect how it happens, then to reproduce, then write a
261 So from the logfile this is what happens: we have a good interval with
262 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
263 top of it. Locally we now mirror F2 and open a new done interval for it.
264 Then we mirror F1 but this time with the timestamp T1b. And when we then
265 try to close the gap, we do not find T1 but instead something older. We
266 should gladly accept this older piece and this would fix this bug.
270 * bug to fix: when the 1h file changes while rmirror is running, we do
271 correctly sync the new files but never switch to the 6h file but rather
272 stay in a rather quick loop that fetches the 1h file again and again.
274 Is it possible that we initialize a new object? Or does
275 get_remote_recentfile_as_tempfile overwrite something in myself?
277 Want a new option: _runstatusfile => $file which frequently dumps the
278 state of all recentfiles to a file.
282 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
284 * Todo: now teach update to verify the timestamp is about to write
285 against the previous and use _increase_a_bit if it doesn't comply with
286 strict monotony. DONE
288 * The problem of rounding. So far perl's default precision was
289 sufficient. One day it won't be. FakeFloat has an easy job when it is
290 only reading and other machines have written correctly. But when we want
291 to write a floating point number that is a bit larger than the other
292 one, then we need our own idea of precision.
294 Slaven said: just append a "1". This might be going towards the end of
295 usability too quickly. I'd like something that actually uses the decimal
296 system. Well, appending a 1 also does this but...
298 E.g. we have 1.0. nextup on this architecture is starting with
299 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
300 taken the 1.0000000000000003 and the next user comes and the time tells
301 him 1.0 again. He has to beat my number without stepping over the
302 nextup. This is much less space than I had when I chose 1,2,3.
304 What is also irritating is that nextup is architecture dependent. The
305 128 bit guy must choose very long numbers to fit in between whereas the
306 other one with 16 bit uses larger steps. But then the algorithm is the
307 same for both, so that would be a nice thing.
309 I see two situation where we need this. One is when Time::HiRes returns
310 us a value that is <= the last entry in our recentfile. In this case
311 (let's call it the end-case) we must fill the region between that number
312 and the next higher native floating point number. The other is when we
313 inject an old file into an old recentfile (we would then also set a new
314 dirtymark). We find the integer value already taken and need a slightly
315 different one (let's call it the middle-case). The difference between
316 the two situations is that the next user will want to find something
317 higher than my number in the end-case and something lower than my number
320 So I suggest we give the function both a value and an upper bound and it
321 calculates us a primitive middle. The upper bound in the middle-case is
322 the next integer. The upper bound on the end-case is the nextup floating
323 point number. But the latter poses another problem: if we have occupied
324 the middle m between x and nextup(x), then the nextup(m) will probably
325 not be the same as nextup(x) because some rounding will take place
326 before the nextup is calculated and when the rounding reaches the
327 nextup(x), we will end up at nextup(nextup(x)).
329 So we really need to consider the nextup and the nextdown from there and
330 then the middle and that's the number we may approach asymptotically.
333 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
335 * consider deprecating the use of RECENT.recent as a symlink. It turns
336 out to need extra hoops with the rsync options and just isn't worth it.
337 Or maybe these extra hoops are needed anyway for the rest of the tree?
338 Nope, can't be the case because not all filesystems support symlinks.
340 But before doing the large step, I'll deprecate the call of
341 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
342 only introduced to resolve RECENT.recent and complicates the routine far
343 beyond what it deserves.
345 DONE. Won't deprecate RECENT.recent, just moved its handling to the
348 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
350 * I think it's a bug that the rsync_option links must be set to true in
351 order to support RECENT.recent and that nobody cares to set it
352 automatically. Similar for ignore_link_stat_errors.
354 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
356 * Todo: find all todos together and make a plan what is missing for a
359 - verifytree or something like that. fsck maybe.
361 - rersyncrecent, the script itself? What it do?
363 - a way to only mirror the recentfiles without mirroring the whole
364 remote system such that people can decide to mirror only partially see
365 also 2008-08-30. .shadow-xxx directory? this also needed for a
366 filesystem that is still incomplete and might need the mirrorfiles for
369 - long living objects that mirror again and again. Inject something
370 into ta, see how it goes over to tb.
372 - how do we continue filling up the DONE system when we use an object
373 for the second time? "fully covered" and "uptodate" or new terminology.
375 - overview called on the wrong file should be understandable
377 - the meta data field that must change when we fake something up so that
378 the downstream people know they have to re-fetch everything.
380 - how tolerant are we against missing files upstream? how do we keep
381 track? there are legitimate cases where we did read upstream index right
382 before a file got deleted there and then find that file as new and want
383 it. There are other cases that are not self healing and must be tracked
386 - how, exactly, do we have to deal with deletes? With rsync errors?
388 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
389 authors) failed: No such file or directory (2)
391 The file above is a delete in 1h and a new in file 1M and the
392 delete in the locally running rmirror did not get propagated to the 1M
393 object. Bug. And the consequence is a standstill.
395 It seems that a slave that works with a file below the principal needs
396 to merge things all the way up to get rid of later deletes. Or keep
397 track of all deletes and skip them later. So we need a trackdeletes.pm
398 similar to the done.pm?
400 see also 2008-08-20 about spurious deletes that really have no add
401 counterpart and yet they are not wrong.
403 - consider the effect when resyncing the recentfile takes longer than
404 the time per loop. Then we never rsync any file. We need to diagnose
405 that and force an increase of that loop time. But when we later are fast
406 enough again because the net has recovered, then we need to switch back
407 to original parameters. ERm, no, it's enough to keep syncing at least
408 one file before refetching an index file.
410 - remember to verify that no temp files are left lying around and the
413 - status file for not long running jobs that want to track upstream with
416 - revisit all XXX _float areas and study Sub::Exporter DONE
418 - persistent DB even though we just said we do not need it. Just for
419 extended capabilities and time savings when, for example, upstream
420 announces a reset and we get new recentfiles and could then limit
421 ourselves to a subset of files (those that have a changed epoch) in a
422 first pass and would only then do the loop to verify the rest. Or
425 * Todo: aggregate files should know their feed and finding the principal
426 should be done stepwise. (?)
428 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
429 left locks around. DONE
431 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
433 * maybe extend the _overview so that it always says if and where the
434 last file is in the next file and where the next event in the next rf
435 would lie. No, don't like this anymore. REJECT
437 * take the two new redundant tests out again, only the third must
440 * Todo: add a sanity check if the merged structure is really pointing to
441 a different rf and that this different rf is larger. DONE
443 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
445 * now test, if they are overlapping. And test if there is a file in the
446 next rf that would fit into this rf's interval.
448 1h 1222324012.8474 1222322541.7963 0.4086
449 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
450 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
451 1W 1222313218.3626 1221708477.5829 167.9835
453 I suspect that somebody writes a merged timestamp without having merged
454 and then somebody else relies on it.
456 If aggregate is running, the intervals must not be extravagated, if it
457 is not running, there must not be bounds, the total number of events in
458 the system must be counted and must be controlled throughout the tests.
459 That the test required the additional update was probably nonsense,
460 because aggregate can cut pieces too. FIXED & DONE
462 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
464 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
467 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
469 * Most apparent bug at the moment is that the recentfiles are fetched
470 too often. Only the principal should be fetched and if it has not
471 changed, the others should not be refetched. ATM I must admit that I'm
472 happy that we refetch more often than needed because I can more easily
473 fix bugs while the thing is running.
475 * Let's say, 1220474966.19501 is a timestamp of a file that is already
476 done but the done system does not know about it. The reason for the
477 failure is not known and we never reach the status uptodate because of
478 this. We must get over it.
480 Later it turns out that the origin server had a bug somewhere.
481 1220474966.19042 came after 1220474966.19501. Or better: it was in the
482 array of the recentfile one position above. The bug was my own.
484 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
486 * There is the race condition where the server does a delete and the
487 slave does not yet know and then tries to download it because he sees
488 the new. So for this time window we must be more tolerant against
489 failure. If we cannot download a file, we should just skip it and should
490 not retry immediately. The whole system should discover the lost thing
491 later. Keeping track with the DONE system should really be a no brainer.
493 But there is something more: the whole filesystem is a database and the
494 recentfiles are one possible representation of it. It's a pretty useful
495 representation I think that's why I have implemented something around
496 it. But for strictly local operation it has little value. For local
497 operation we would much rather have a database. So we would enter every
498 recentfile reading and every rsync operation and for every file the last
499 state change and what it leads to. Then we would always ignore older
500 records without the efforts involved with recentfiles.
502 The database would have: path,recentepoch,rsyncedon,deletedon
504 Oh well, not yet clear where this leads to.
506 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
508 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
509 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
511 * statusfile, maybe only on demand, alone to have a sharp debugging
512 tool. It is locked and all recentfiles dump themselves into it and we
513 can build a viewer that lets us know where we stand and what's inside.
515 * remember: only the principal recentfile needs expiration, all others
516 shall be expired by principal if it discovers that something has move
519 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
521 * Always check if we stringify to a higher value than in the entry
524 * And in covered make an additional check if we would be able to see a
525 numerical difference between the two numbers and if we can't then switch
526 to a different, more expensive algorithm. Do not want to be caught by
527 floating surprises. DONE
529 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
531 * caching has several aspects here: we can cache the interval of the
532 recentfile which only will change when the mtime of the file changes. We
533 must re-mirror the recentfile when its ttl has expired. Does have_read
534 tell you anything? It counts nothing at all. Only the mtime is
535 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
536 thing the have_mirrored because it is unrelated to the mtime.
538 * Robustness of floating point calculations! I always thought that the
539 string calculated by the origin server for the floating representation
540 of the epoch time is just a string. When we convert it to a number and
541 later back to a string, the other computer might come to a different
542 conclusion. This must not happen, we want to preserve it under any
543 circumstances. I will have to write tests with overlong sequences that
544 get lost in arithmetic and must see if all still works well. DONE
546 But one fragile point remains: if one host considers a>b and the other
547 one considers them == but no eq. To prevent this, we must probably do
548 some extra homework. DONE
550 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
552 * the concept of tracking DONE needs an object per recentfile that has
553 something like these methods:
555 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
557 covered() register() covered()
559 The unclear thing is how we translate points in time into intervals. We
560 could pass a reference to the current recent_events array when running
561 we_have(xxx) and let the DONE object iterate over it such that it only
562 has to store a list of intervals that can melt into each other. Ah, even
563 passing the list together with a list of indexes seems feasiable.
565 Or maybe ask for the inverted list?
567 Whenever the complete array is covered by the interval we say we are
568 fully covered and if the recentfile is not expired, we are uptodate.
570 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
572 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
574 * need a way to "return" the next entry after the end of a list. When
575 the caller says "before" or "after" we would like to know if he could
576 cover that interval/threshold or not because this influences the effect
577 of a newer timestamp of that recentfile. DONE with $opt{info}.
579 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
581 * one of the next things to tackle: the equivalent of csync2 -TIXU.
583 loop implies tixu (?). Nope, something like --statefile decides. Per
586 T test, I init, X including removals, U nodirtymark
588 So we have no concept of dirtymarks, we only trust that since we are
589 running we have observed everything steadily. But people will not let
590 this program run forever so we must consider both startup penalty and
591 book keeping for later runs. We keep this for later. For now we write a
592 long running mirror that merges several intervals.
594 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
596 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
599 * rersyncrecent, the script: default to one week. The name of the switch
600 is --after. Other switches? --loop!
602 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
604 * need a switch --skip-deletes (?)
606 * need a switch --enduser that tells us that the whole tempfile
607 discipline is not needed when there is no downstream user. (?)
609 Without this switch we cannot have a reasonable recent.pl that just
610 displays the recent additions. Either we accept to download everything.
611 Or we download temporary files without the typical rsync protocol
614 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
615 File::Temp::tempdir, this might be a win.
617 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
619 * apropos missing: we have no push, we never know the downstream
620 servers. People who know their downstream hosts and want to ascertain
621 something will want additional methods we have never thought about, like
622 update or delete a certain file.
624 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
626 * tempted to refactor rmirror into resolve_symlink, localize, etc.
627 Curious if rsync_options=links equal 0 vs. 1 will make the expected
630 * rsync options: it's a bit of a pain that we usually need several rsync
631 options, like compress, links, times, checksum and that there is no
632 reasonable default except the original rsync default. I think wee can
633 safely assume that the rsync options are shared between all recentfile
634 instances within one recent tree.
636 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
638 * deletes: if a delete follows an add quickly enough it may happen that
639 a downstream mirror did not see the add at all! It seems this needs to
640 be mentioned somewhere. The point here is that even if the downstream is
641 never missing the principal timeframe it may encounter a "delete" that
642 has no complimentary "add" anywhere.
644 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
646 * I suspect the treat of metadata is incorrect during read or something.
647 The bug that I am watching is that between 06:08 and 06:09 the 6h file
648 contained more than 6 hours worth of data. At 06:08 we merged into the
649 1d file. We need to take snapshots of the 6h file over the course of an
650 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
653 Much worse: watching the 1h file: right at the moment (at 06:35) it
654 covers 1218867584-1219120397 which is 70 hours.
656 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
657 08:19:44 2008, that is when I checked out last time, so it seems to be
658 aggregating and never truncating?
660 No, correct is: it is never truncating; but wrong is: it is aggregating.
661 It does receive a lot of events from time to time from a larger file.
662 Somehow a large file gets merged into the small one and because the
663 "meta/merged" attribute is missing, nobody is paying attention. I
664 believe that I can fix this by making sure that metadata are honoured
665 during read. DONE and test adjusted.
667 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
669 * grand renaming plan
671 remotebase => remoteroot to fit well with localroot DONE
672 local_path() => localroot seems to me should already work DONE
673 recentfile_basename => rfilename no need to stress it has no slash DONE
675 filenameroot??? Doesn't seem too bad to me today. Maybe something like
676 kern? It would anyway need a deprecation cycle because it is an
677 important constructor.
679 * I like the portability that Data::Serializer brings us but the price
680 is that some day we might find out that it is slowing us a bit. We'll
683 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
685 * should we not enter the interval of the principal (or the interval of
686 the merging file?) in every aggregated/merged file?
688 * we should aim at a first release and give up on thinking about
689 sanitizing stuff and zloop. Let's just admit that a full traditional
690 rsync is the only available sanitizer ATM. Otherwise it's complicated
691 stuff: sanitizing on the origin server, sanitizing on the slaves,
692 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
693 get the basics out before this becomes a major cause for mess.
695 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
697 * On OSes not supporting symlinks we expect that RECENT.recent contains
698 the contents of the principal recentfile. Actually this is identical on
699 systems supporting symlinks. Simple, what follows from that is that we
700 need to keep the serializer in the metadata because we cannot read it
701 from the filename, doesn't it? Of course not. It's a chicken and egg
702 problem. This leaves us with the problem to actually parse the
703 serialized data to find out in which format it is. So who can do the 4
704 or 5 magics we wanted to support? File::LibMagic?
706 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
708 * remotebase and recentfile_basename are ugly names. Now that we need a
709 word for the shortest/principal/driving recentfile too we should do
712 localroot is good. rfile is good. local_path() is bad, local_path($path)
713 is medium, filenameroot() is bad, remotebase is bad, recentfile is
716 Up to now remotebase was the string that described the remote root
717 directory in rsync notation, like pause.perl.org::authors. And
718 recentfile_basename was "RECENT-1h.yaml".
720 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
722 * The test that was added in today's checkin is a good start for a test
723 of rmirror. We should have more methods in Recent.pm: verify,
724 addmissingfiles. We should verify the current tree, then rmirror it and
725 then verifytree the copy. We could then add some arbitrary file and let
726 it be discovered by addmissingfiles, then rmirror again and then
727 verifytree the copy again.
729 Then we could start stealing from csync2 sqlite database [no port to
730 OSX!] and fill a local DB. And methods to compare the database with the
731 recentfiles. Our strength is that in principle we could maintain state
732 with a single float. We have synced up to 1234567890.123456. If the Z
733 file does not add new files all we have to do is mirror the new ones and
736 This makes it clear that we should extend current protocol and declare
737 that we cheat when we add files too late, just to help the other end
738 keeping track. Ah yes, that's what was meant when zloop was mentioned
741 Maybe need to revisit File::Mirror to help me with this task.
743 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
745 * There must be an allow-me-to-truncate flag in every recentfile.
746 Without it one could construct a sequence of updates winning the locking
747 battle against the aggregator. Only if an aggregator has managed to
748 merge data over to the next level, truncating can be allowed. DONE with
751 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
753 * We should probably guarantee that no duplicates enter the aggregator
756 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
758 * To get merge operation faster would need a good benchmark test. What
759 02 spits out isn't reliable enough and is dominated by many other
762 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
763 Author: Andreas J. Koenig <k@k75.(none)>
764 Date: Sat Aug 2 07:58:04 2008 +0200
768 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
769 Author: Andreas J. Koenig <k@k75.(none)>
770 Date: Sat Aug 2 11:40:29 2008 +0200
772 there was a lot of trying but the effect is hardly measurable with
775 * overhead of connecting seems high. When setting
776 max_files_per_connection to 1 we see that.
778 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
780 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
782 25138.2215600014/3600 = 6.98283932222261
784 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
785 be a bug in the aggregator. FIXED
787 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
789 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
790 we have it, can we mirror it, mirror it!
792 I fear this needs a new class which might be called
793 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
794 and treat them as an entity. I realize that a single recentfile may be
795 sufficient for certain tasks and that it is handy for the low level
796 programmer but it is not nice to use. If there is a delete in the 1h
797 file then the 6h file still contains it. Seekers of the best information
798 need to combine at least some of the recentfiles most of the time.
800 There is the place for the Z loop!
802 But the combination is something to collect in a database, isn't it. Did
803 csync2 just harrumph?
805 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
807 * it just occurred to me that hosts in the same mirroring pool could
808 help out each other even without rewriting the recentfile. Just fetch
809 the stuff to mirror from several places, bingo. But that's something
810 that should rather live in a separate package or in rsync directly.
812 * cronjobs are unsuited because with ntp they would all come at the full
813 minute and disturb each other. Besides that I'd hate to have a backbone
814 with more than a few seconds latency.
816 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
818 * a second rsync server with access control for PAUSE. Port? 873 is the
819 standard port, let's take 8873.
821 * if there were a filesystem based on this, it would have a slow access
822 to inexistent files. It would probably provide wrong readdir (only based
823 on current content) or also a slow one (based on a recentfile written
824 after the call). But it would provide fast access to existing files. Or
825 one would deliberately allow slightly blurred answers based on some
826 sqlite reflection of the recentfiles.
828 * todo: write a variant of mirror() that combines two or more
829 recentfiles and treats them like one
831 * todo: signal handler to remove the tempfile
833 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
835 * now that we have the symlink I forgot how it should be used in
838 * the z loop: add missing files to Z file. Just append them (instead of
839 prepending). So one guy prepends something from the Y file from time to
840 time and another guy appends something rather frequently. Collecting
841 pond. When Y merges into Z, things get epoch and the collecting pond
842 gets smaller. What exactly are "missing files"?
844 take note of current epoch of the alpha file, let's call it the
847 find all files on disk
849 remove all files registered in the recentworld up to recent-ts
851 remove all files that have been deleted after recent-ts according to
854 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
856 * rersyncrecent might be a cronjob with a (locked) state file which
857 contains things like after and maybe last z sync or such?
859 rrr-mirror might be an alternative name but how would we justify the
860 three Rs when there is no Re-Rsync-Recent?
862 With the --loop parameter it is an endless loop, without it is no loop.
863 At least this is simple.
865 * todo: new accssor z-interval specifies how often the Z file is updated
866 against the filesystem. We probably want no epoch stamp on these
867 entries. And we want to be able to filter the entries (e.g. no
868 by-modules and by-category tree)
870 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
872 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
873 into Z. We do not need the other files filled up with filesystem
876 * need interface to query for a file in order to NOT call update on
877 PAUSE a second time within a short time.
879 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
881 * recommended update interval? Makes no sense, is different for
888 change-log-default-name: "Todo"