1 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * memory leak in the syncher? It currently weighs 100M.
7 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
10 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
12 * lookup by epoch and by path and use this ability on the pause to never
13 again register a file twice that doesn't need it. Let's call it
16 * after the dirtymark is done: fill up recentfiles with fake (historic)
17 entries; fill up with individual corrections; algorithm maybe to be done
18 with bigfloat so that we can always place something in the middle
19 between two entries. Before we must switch to bigfloat we could try to
20 use Data::Float::nextup to get the.
22 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
25 * dirtymark now lives in Recentfile, needs to be used in rmirror.
27 * find out why the downloader died after a couple of hours without a net
28 connection. Write a test that survives the not-existence of the other
31 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
33 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
34 it makes less sense with tight coupling for secondary files. But asking
35 the server what the current epoch is might be cheaper on HTTP than on
36 rsync. (Needs to be evaluated)
38 * remove the 0.00 from the verbose overview in the Merged column in the
41 * write tests that expose the problems of the last few days: cascading
42 client/server roles, tight coupling for secondary RFs, deletes after
45 * Some day we might want to have policy options for the slave:
46 tight/loose/no coupling with upstream for secondary RFs. tight is what
47 we have now. loose would wait until a gap occurs that can be closed.
49 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
51 * revisit all $rfs->[$i+1] places if they now make sense still
53 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
55 * another bug is the fact that the mirror command deletes files before
56 it unhides the index file, thus confusing downstream slaves. We must not
57 delete before unhiding and must delete after unhiding. FIXED.
59 * new complication about the slave that is playing a server role.
60 Currently we mirror from newest to oldest with a hidden temporary file
61 as index. And when one file is finished, we unhide the index file.
62 Imagine the cascading server/slave is dead for a day. It then starts
63 mirroring again with the freshest thing and unhides the freshest index
64 file when it has worked through it. In that moment it exposes a time
65 hole. Because it now works on the second recentfile which is still
68 We currently do nothing special to converge after such a drop out. At
69 least not intentionally and robustly and thought through.
71 The algorithm we use to seed the next file needs quite a lot of more
72 robustness than it currently has. Something to do with looking at the
73 merged element of the next rf and when it has dropped off, we seed
74 immediately. And if it ramains dropped off, we seed again, of course.
76 Nope, looking from smaller to larger RFS we look at the merged element
77 of this RF and at the minmax/max element of the next RF. If that
78 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
81 And we need a public accessor seed and unseed or seeded. But not the mix
82 of public and private stuff that then is used behind the back.
84 And then the secondary* stuff must go.
86 And we must understand what the impact is on the DONE system. Can it go
87 unnoticed that there was a hole? And could the DONE system have decided
88 the hole is covered? This should be testable with three directories where
89 the middle stops working for a while. Done->merge is suspicious, we must
90 stop it from merging non-conflatable neighbors due to broken continuity.
94 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
96 * Slaven suggests to have the current epoch or the whole current
97 recentfile available from the HTTP server and take it away with
98 keepalive. This direction goes the granularity down to subseconds.
100 We might want to rewrite everything to factor out transport and allow
101 the whole thing to run via HTTP.
103 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
105 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
108 * smoker on k81 fetching from k75 to verify cascading works. See
109 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
111 * maybe the loop should wait for CHECKSUMS file after every upload. And
112 CPAN.pm needs to deal with timestamps in the future.
114 * do not forget the dirtymark!
116 Text: have a new flag on recentfiles with the meaning: if this
117 changes, you're required to run a full rsync over all the files. The
118 reason why we set it would probably be: some foul happened. we injected
119 files in arbitrary places or didn't inject them although they changed.
120 The content of the flag? Timestamp? The relation between the
121 recentfiles would have to be inheritance from the principal, because any
122 out of band changes would soon later propagate to the next recentfile.
124 By upping the flag often one can easily ruin the slaves.
126 last out of band change? dirtymark?
128 Anyway, this implies that we read a potentially existing recentfile
131 And it implies that we have an eventloop that keeps us busy in 2-3
132 cycles, one for current stuff (tight loop) and one for the recentfiles
133 (cascade when principal has changed), one for the old stuff after a
136 And it implies that the out-of-band change in any of the recentfiles
137 must have a lock on the principal file and there is the place to set the
140 * start a FAQ, especially quick start guide questions. Also to aid those
141 problematic areas where we have no good solution, like the "links"
144 * wish feedback when we are slow.
148 * Remove a few DEBUG statements.
150 * The multiple-rrr way of doing things needs a new option to rmirror,
151 like piecemeal or so. Not urgent because after the first pass through,
152 things run smoothely. It's only ugly during the first pass.
154 * I have the suspicion that the code is broken that decides if the
155 neighboring RF needs to be seeded. I fear when too much time has gone
156 between two calls (in our case more than one hour), it would not seed
157 the neighbor. Of course this will never be noticed, so we need a good
160 * local/localroot confusion: I currently pass both options but one must
163 * accounts for early birds on PAUSE rsync daemon.
165 * hardcoded 20 seconds
167 * who mirrors the index? DOING now.
169 * which CPAN mirrors offer rsync?
171 * visit all XXX, visit all _float places
173 * rename the pathdb stuff, it's too confusing. No idea how.
175 * rrr-inotify, backpan, rrr-register
177 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
179 * current bugs: the pathdb seems to get no reset, the seeding of the
180 secondaryttl stuff seems not to have an effect. Have helped myself with
181 a rand(10), need to fix this back. So not checked in. Does the rand
184 The rand thing helps. The secondaryttl stuff was in the wrong line,
187 The pathdb stuff was because I called either _pathdb or __pathdb on the
188 wrong object. FIXED now.
190 * It's not so beautiful if we never fetch the recentfiles that are not
191 the principal, even if this is correct behaviour. We really do not need
192 them after we have fetched the whole content.
194 OK, we want a switch for that: secondaryttl DONE
196 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
198 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
199 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
200 it lets 40 follow. FIXED
202 * See that the long running process really only updates the principal
203 file unless it has missed a timespan during which something happened. If
204 nothing happened, it must notice even when it misses the timespan. DONE
206 * we must throw away the pathdb when we have reached the end of Z. From
207 that moment we can have a very small pathdb because the only reason for
208 a pathdb is that we know to ignore old records in old files. We won't
209 need this pathdb again before the next full pass over the data is
210 necessary and then we will rebuild it as we go along. DONE
212 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
214 * I think, Done::register_one is doing wrong in that it does not
215 conflate neighboring pieces. The covered() method cannot do this because
216 it has no recent_events array at hand. But register_one has it and could
217 do it and for some reason misses to do it (sometimes).
219 This means that the three tests I just wrote can probably not survive
220 because they test with an already broken Done structure.
222 The art now is to detect how it happens, then to reproduce, then write a
225 So from the logfile this is what happens: we have a good interval with
226 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
227 top of it. Locally we now mirror F2 and open a new done interval for it.
228 Then we mirror F1 but this time with the timestamp T1b. And when we then
229 try to close the gap, we do not find T1 but instead something older. We
230 should gladly accept this older piece and this would fix this bug.
234 * bug to fix: when the 1h file changes while rmirror is running, we do
235 correctly sync the new files but never switch to the 6h file but rather
236 stay in a rather quick loop that fetches the 1h file again and again.
238 Is it possible that we initialize a new object? Or does
239 get_remote_recentfile_as_tempfile overwrite something in myself?
241 Want a new option: _runstatusfile => $file which frequently dumps the
242 state of all recentfiles to a file.
246 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
248 * Todo: now teach update to verify the timestamp is about to write
249 against the previous and use _increase_a_bit if it doesn't comply with
250 strict monotony. DONE
252 * The problem of rounding. So far perl's default precision was
253 sufficient. One day it won't be. FakeFloat has an easy job when it is
254 only reading and other machines have written correctly. But when we want
255 to write a floating point number that is a bit larger than the other
256 one, then we need our own idea of precision.
258 Slaven said: just append a "1". This might be going towards the end of
259 usability too quickly. I'd like something that actually uses the decimal
260 system. Well, appending a 1 also does this but...
262 E.g. we have 1.0. nextup on this architecture is starting with
263 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
264 taken the 1.0000000000000003 and the next user comes and the time tells
265 him 1.0 again. He has to beat my number without stepping over the
266 nextup. This is much less space than I had when I chose 1,2,3.
268 What is also irritating is that nextup is architecture dependent. The
269 128 bit guy must choose very long numbers to fit in between whereas the
270 other one with 16 bit uses larger steps. But then the algorithm is the
271 same for both, so that would be a nice thing.
273 I see two situation where we need this. One is when Time::HiRes returns
274 us a value that is <= the last entry in our recentfile. In this case
275 (let's call it the end-case) we must fill the region between that number
276 and the next higher native floating point number. The other is when we
277 inject an old file into an old recentfile (we would then also set a new
278 dirtymark). We find the integer value already taken and need a slightly
279 different one (let's call it the middle-case). The difference between
280 the two situations is that the next user will want to find something
281 higher than my number in the end-case and something lower than my number
284 So I suggest we give the function both a value and an upper bound and it
285 calculates us a primitive middle. The upper bound in the middle-case is
286 the next integer. The upper bound on the end-case is the nextup floating
287 point number. But the latter poses another problem: if we have occupied
288 the middle m between x and nextup(x), then the nextup(m) will probably
289 not be the same as nextup(x) because some rounding will take place
290 before the nextup is calculated and when the rounding reaches the
291 nextup(x), we will end up at nextup(nextup(x)).
293 So we really need to consider the nextup and the nextdown from there and
294 then the middle and that's the number we may approach asymptotically.
297 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
299 * consider deprecating the use of RECENT.recent as a symlink. It turns
300 out to need extra hoops with the rsync options and just isn't worth it.
301 Or maybe these extra hoops are needed anyway for the rest of the tree?
302 Nope, can't be the case because not all filesystems support symlinks.
304 But before doing the large step, I'll deprecate the call of
305 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
306 only introduced to resolve RECENT.recent and complicates the routine far
307 beyond what it deserves.
309 DONE. Won't deprecate RECENT.recent, just moved its handling to the
312 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
314 * I think it's a bug that the rsync_option links must be set to true in
315 order to support RECENT.recent and that nobody cares to set it
316 automatically. Similar for ignore_link_stat_errors.
318 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
320 * Todo: find all todos together and make a plan what is missing for a
323 - verifytree or something like that. fsck maybe.
325 - rersyncrecent, the script itself? What it do?
327 - a way to only mirror the recentfiles without mirroring the whole
328 remote system such that people can decide to mirror only partially see
329 also 2008-08-30. .shadow-xxx directory? this also needed for a
330 filesystem that is still incomplete and might need the mirrorfiles for
333 - long living objects that mirror again and again. Inject something
334 into ta, see how it goes over to tb.
336 - how do we continue filling up the DONE system when we use an object
337 for the second time? "fully covered" and "uptodate" or new terminology.
339 - overview called on the wrong file should be understandable
341 - the meta data field that must change when we fake something up so that
342 the downstream people know they have to re-fetch everything.
344 - how tolerant are we against missing files upstream? how do we keep
345 track? there are legitimate cases where we did read upstream index right
346 before a file got deleted there and then find that file as new and want
347 it. There are other cases that are not self healing and must be tracked
350 - how, exactly, do we have to deal with deletes? With rsync errors?
352 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
353 authors) failed: No such file or directory (2)
355 The file above is a delete in 1h and a new in file 1M and the
356 delete in the locally running rmirror did not get propagated to the 1M
357 object. Bug. And the consequence is a standstill.
359 It seems that a slave that works with a file below the principal needs
360 to merge things all the way up to get rid of later deletes. Or keep
361 track of all deletes and skip them later. So we need a trackdeletes.pm
362 similar to the done.pm?
364 see also 2008-08-20 about spurious deletes that really have no add
365 counterpart and yet they are not wrong.
367 - consider the effect when resyncing the recentfile takes longer than
368 the time per loop. Then we never rsync any file. We need to diagnose
369 that and force an increase of that loop time. But when we later are fast
370 enough again because the net has recovered, then we need to switch back
371 to original parameters. ERm, no, it's enough to keep syncing at least
372 one file before refetching an index file.
374 - remember to verify that no temp files are left lying around and the
377 - status file for not long running jobs that want to track upstream with
380 - revisit all XXX _float areas and study Sub::Exporter DONE
382 - persistent DB even though we just said we do not need it. Just for
383 extended capabilities and time savings when, for example, upstream
384 announces a reset and we get new recentfiles and could then limit
385 ourselves to a subset of files (those that have a changed epoch) in a
386 first pass and would only then do the loop to verify the rest. Or
389 * Todo: aggregate files should know their feed and finding the principal
390 should be done stepwise. (?)
392 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
393 left locks around. DONE
395 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
397 * maybe extend the _overview so that it always says if and where the
398 last file is in the next file and where the next event in the next rf
399 would lie. No, don't like this anymore. REJECT
401 * take the two new redundant tests out again, only the third must
404 * Todo: add a sanity check if the merged structure is really pointing to
405 a different rf and that this different rf is larger. DONE
407 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
409 * now test, if they are overlapping. And test if there is a file in the
410 next rf that would fit into this rf's interval.
412 1h 1222324012.8474 1222322541.7963 0.4086
413 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
414 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
415 1W 1222313218.3626 1221708477.5829 167.9835
417 I suspect that somebody writes a merged timestamp without having merged
418 and then somebody else relies on it.
420 If aggregate is running, the intervals must not be extravagated, if it
421 is not running, there must not be bounds, the total number of events in
422 the system must be counted and must be controlled throughout the tests.
423 That the test required the additional update was probably nonsense,
424 because aggregate can cut pieces too. FIXED & DONE
426 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
428 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
431 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
433 * Most apparent bug at the moment is that the recentfiles are fetched
434 too often. Only the principal should be fetched and if it has not
435 changed, the others should not be refetched. ATM I must admit that I'm
436 happy that we refetch more often than needed because I can more easily
437 fix bugs while the thing is running.
439 * Let's say, 1220474966.19501 is a timestamp of a file that is already
440 done but the done system does not know about it. The reason for the
441 failure is not known and we never reach the status uptodate because of
442 this. We must get over it.
444 Later it turns out that the origin server had a bug somewhere.
445 1220474966.19042 came after 1220474966.19501. Or better: it was in the
446 array of the recentfile one position above. The bug was my own.
448 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
450 * There is the race condition where the server does a delete and the
451 slave does not yet know and then tries to download it because he sees
452 the new. So for this time window we must be more tolerant against
453 failure. If we cannot download a file, we should just skip it and should
454 not retry immediately. The whole system should discover the lost thing
455 later. Keeping track with the DONE system should really be a no brainer.
457 But there is something more: the whole filesystem is a database and the
458 recentfiles are one possible representation of it. It's a pretty useful
459 representation I think that's why I have implemented something around
460 it. But for strictly local operation it has little value. For local
461 operation we would much rather have a database. So we would enter every
462 recentfile reading and every rsync operation and for every file the last
463 state change and what it leads to. Then we would always ignore older
464 records without the efforts involved with recentfiles.
466 The database would have: path,recentepoch,rsyncedon,deletedon
468 Oh well, not yet clear where this leads to.
470 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
472 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
473 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
475 * statusfile, maybe only on demand, alone to have a sharp debugging
476 tool. It is locked and all recentfiles dump themselves into it and we
477 can build a viewer that lets us know where we stand and what's inside.
479 * remember: only the principal recentfile needs expiration, all others
480 shall be expired by principal if it discovers that something has move
483 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
485 * Always check if we stringify to a higher value than in the entry
488 * And in covered make an additional check if we would be able to see a
489 numerical difference between the two numbers and if we can't then switch
490 to a different, more expensive algorithm. Do not want to be caught by
491 floating surprises. DONE
493 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
495 * caching has several aspects here: we can cache the interval of the
496 recentfile which only will change when the mtime of the file changes. We
497 must re-mirror the recentfile when its ttl has expired. Does have_read
498 tell you anything? It counts nothing at all. Only the mtime is
499 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
500 thing the have_mirrored because it is unrelated to the mtime.
502 * Robustness of floating point calculations! I always thought that the
503 string calculated by the origin server for the floating representation
504 of the epoch time is just a string. When we convert it to a number and
505 later back to a string, the other computer might come to a different
506 conclusion. This must not happen, we want to preserve it under any
507 circumstances. I will have to write tests with overlong sequences that
508 get lost in arithmetic and must see if all still works well. DONE
510 But one fragile point remains: if one host considers a>b and the other
511 one considers them == but no eq. To prevent this, we must probably do
512 some extra homework. DONE
514 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
516 * the concept of tracking DONE needs an object per recentfile that has
517 something like these methods:
519 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
521 covered() register() covered()
523 The unclear thing is how we translate points in time into intervals. We
524 could pass a reference to the current recent_events array when running
525 we_have(xxx) and let the DONE object iterate over it such that it only
526 has to store a list of intervals that can melt into each other. Ah, even
527 passing the list together with a list of indexes seems feasiable.
529 Or maybe ask for the inverted list?
531 Whenever the complete array is covered by the interval we say we are
532 fully covered and if the recentfile is not expired, we are uptodate.
534 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
536 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
538 * need a way to "return" the next entry after the end of a list. When
539 the caller says "before" or "after" we would like to know if he could
540 cover that interval/threshold or not because this influences the effect
541 of a newer timestamp of that recentfile. DONE with $opt{info}.
543 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
545 * one of the next things to tackle: the equivalent of csync2 -TIXU.
547 loop implies tixu (?). Nope, something like --statefile decides. Per
550 T test, I init, X including removals, U nodirtymark
552 So we have no concept of dirtymarks, we only trust that since we are
553 running we have observed everything steadily. But people will not let
554 this program run forever so we must consider both startup penalty and
555 book keeping for later runs. We keep this for later. For now we write a
556 long running mirror that merges several intervals.
558 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
560 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
563 * rersyncrecent, the script: default to one week. The name of the switch
564 is --after. Other switches? --loop!
566 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
568 * need a switch --skip-deletes (?)
570 * need a switch --enduser that tells us that the whole tempfile
571 discipline is not needed when there is no downstream user. (?)
573 Without this switch we cannot have a reasonable recent.pl that just
574 displays the recent additions. Either we accept to download everything.
575 Or we download temporary files without the typical rsync protocol
578 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
579 File::Temp::tempdir, this might be a win.
581 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
583 * apropos missing: we have no push, we never know the downstream
584 servers. People who know their downstream hosts and want to ascertain
585 something will want additional methods we have never thought about, like
586 update or delete a certain file.
588 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
590 * tempted to refactor rmirror into resolve_symlink, localize, etc.
591 Curious if rsync_options=links equal 0 vs. 1 will make the expected
594 * rsync options: it's a bit of a pain that we usually need several rsync
595 options, like compress, links, times, checksum and that there is no
596 reasonable default except the original rsync default. I think wee can
597 safely assume that the rsync options are shared between all recentfile
598 instances within one recent tree.
600 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
602 * deletes: if a delete follows an add quickly enough it may happen that
603 a downstream mirror did not see the add at all! It seems this needs to
604 be mentioned somewhere. The point here is that even if the downstream is
605 never missing the principal timeframe it may encounter a "delete" that
606 has no complimentary "add" anywhere.
608 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
610 * I suspect the treat of metadata is incorrect during read or something.
611 The bug that I am watching is that between 06:08 and 06:09 the 6h file
612 contained more than 6 hours worth of data. At 06:08 we merged into the
613 1d file. We need to take snapshots of the 6h file over the course of an
614 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
617 Much worse: watching the 1h file: right at the moment (at 06:35) it
618 covers 1218867584-1219120397 which is 70 hours.
620 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
621 08:19:44 2008, that is when I checked out last time, so it seems to be
622 aggregating and never truncating?
624 No, correct is: it is never truncating; but wrong is: it is aggregating.
625 It does receive a lot of events from time to time from a larger file.
626 Somehow a large file gets merged into the small one and because the
627 "meta/merged" attribute is missing, nobody is paying attention. I
628 believe that I can fix this by making sure that metadata are honoured
629 during read. DONE and test adjusted.
631 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
633 * grand renaming plan
635 remotebase => remoteroot to fit well with localroot DONE
636 local_path() => localroot seems to me should already work DONE
637 recentfile_basename => rfilename no need to stress it has no slash DONE
639 filenameroot??? Doesn't seem too bad to me today. Maybe something like
640 kern? It would anyway need a deprecation cycle because it is an
641 important constructor.
643 * I like the portability that Data::Serializer brings us but the price
644 is that some day we might find out that it is slowing us a bit. We'll
647 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
649 * should we not enter the interval of the principal (or the interval of
650 the merging file?) in every aggregated/merged file?
652 * we should aim at a first release and give up on thinking about
653 sanitizing stuff and zloop. Let's just admit that a full traditional
654 rsync is the only available sanitizer ATM. Otherwise it's complicated
655 stuff: sanitizing on the origin server, sanitizing on the slaves,
656 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
657 get the basics out before this becomes a major cause for mess.
659 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
661 * On OSes not supporting symlinks we expect that RECENT.recent contains
662 the contents of the principal recentfile. Actually this is identical on
663 systems supporting symlinks. Simple, what follows from that is that we
664 need to keep the serializer in the metadata because we cannot read it
665 from the filename, doesn't it? Of course not. It's a chicken and egg
666 problem. This leaves us with the problem to actually parse the
667 serialized data to find out in which format it is. So who can do the 4
668 or 5 magics we wanted to support? File::LibMagic?
670 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
672 * remotebase and recentfile_basename are ugly names. Now that we need a
673 word for the shortest/principal/driving recentfile too we should do
676 localroot is good. rfile is good. local_path() is bad, local_path($path)
677 is medium, filenameroot() is bad, remotebase is bad, recentfile is
680 Up to now remotebase was the string that described the remote root
681 directory in rsync notation, like pause.perl.org::authors. And
682 recentfile_basename was "RECENT-1h.yaml".
684 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
686 * The test that was added in today's checkin is a good start for a test
687 of rmirror. We should have more methods in Recent.pm: verify,
688 addmissingfiles. We should verify the current tree, then rmirror it and
689 then verifytree the copy. We could then add some arbitrary file and let
690 it be discovered by addmissingfiles, then rmirror again and then
691 verifytree the copy again.
693 Then we could start stealing from csync2 sqlite database [no port to
694 OSX!] and fill a local DB. And methods to compare the database with the
695 recentfiles. Our strength is that in principle we could maintain state
696 with a single float. We have synced up to 1234567890.123456. If the Z
697 file does not add new files all we have to do is mirror the new ones and
700 This makes it clear that we should extend current protocol and declare
701 that we cheat when we add files too late, just to help the other end
702 keeping track. Ah yes, that's what was meant when zloop was mentioned
705 Maybe need to revisit File::Mirror to help me with this task.
707 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
709 * There must be an allow-me-to-truncate flag in every recentfile.
710 Without it one could construct a sequence of updates winning the locking
711 battle against the aggregator. Only if an aggregator has managed to
712 merge data over to the next level, truncating can be allowed. DONE with
715 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
717 * We should probably guarantee that no duplicates enter the aggregator
720 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
722 * To get merge operation faster would need a good benchmark test. What
723 02 spits out isn't reliable enough and is dominated by many other
726 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
727 Author: Andreas J. Koenig <k@k75.(none)>
728 Date: Sat Aug 2 07:58:04 2008 +0200
732 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
733 Author: Andreas J. Koenig <k@k75.(none)>
734 Date: Sat Aug 2 11:40:29 2008 +0200
736 there was a lot of trying but the effect is hardly measurable with
739 * overhead of connecting seems high. When setting
740 max_files_per_connection to 1 we see that.
742 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
744 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
746 25138.2215600014/3600 = 6.98283932222261
748 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
749 be a bug in the aggregator. FIXED
751 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
753 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
754 we have it, can we mirror it, mirror it!
756 I fear this needs a new class which might be called
757 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
758 and treat them as an entity. I realize that a single recentfile may be
759 sufficient for certain tasks and that it is handy for the low level
760 programmer but it is not nice to use. If there is a delete in the 1h
761 file then the 6h file still contains it. Seekers of the best information
762 need to combine at least some of the recentfiles most of the time.
764 There is the place for the Z loop!
766 But the combination is something to collect in a database, isn't it. Did
767 csync2 just harrumph?
769 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
771 * it just occurred to me that hosts in the same mirroring pool could
772 help out each other even without rewriting the recentfile. Just fetch
773 the stuff to mirror from several places, bingo. But that's something
774 that should rather live in a separate package or in rsync directly.
776 * cronjobs are unsuited because with ntp they would all come at the full
777 minute and disturb each other. Besides that I'd hate to have a backbone
778 with more than a few seconds latency.
780 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
782 * a second rsync server with access control for PAUSE. Port? 873 is the
783 standard port, let's take 8873.
785 * if there were a filesystem based on this, it would have a slow access
786 to inexistent files. It would probably provide wrong readdir (only based
787 on current content) or also a slow one (based on a recentfile written
788 after the call). But it would provide fast access to existing files. Or
789 one would deliberately allow slightly blurred answers based on some
790 sqlite reflection of the recentfiles.
792 * todo: write a variant of mirror() that combines two or more
793 recentfiles and treats them like one
795 * todo: signal handler to remove the tempfile
797 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
799 * now that we have the symlink I forgot how it should be used in
802 * the z loop: add missing files to Z file. Just append them (instead of
803 prepending). So one guy prepends something from the Y file from time to
804 time and another guy appends something rather frequently. Collecting
805 pond. When Y merges into Z, things get epoch and the collecting pond
806 gets smaller. What exactly are "missing files"?
808 take note of current epoch of the alpha file, let's call it the
811 find all files on disk
813 remove all files registered in the recentworld up to recent-ts
815 remove all files that have been deleted after recent-ts according to
818 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
820 * rersyncrecent might be a cronjob with a (locked) state file which
821 contains things like after and maybe last z sync or such?
823 rrr-mirror might be an alternative name but how would we justify the
824 three Rs when there is no Re-Rsync-Recent?
826 With the --loop parameter it is an endless loop, without it is no loop.
827 At least this is simple.
829 * todo: new accssor z-interval specifies how often the Z file is updated
830 against the filesystem. We probably want no epoch stamp on these
831 entries. And we want to be able to filter the entries (e.g. no
832 by-modules and by-category tree)
834 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
836 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
837 into Z. We do not need the other files filled up with filesystem
840 * need interface to query for a file in order to NOT call update on
841 PAUSE a second time within a short time.
843 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
845 * recommended update interval? Makes no sense, is different for
852 change-log-default-name: "Todo"