1 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * 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
7 https://rt.cpan.org/Ticket/Display.html?id=41199
9 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
11 Strangest output being HAKANARDO who managed to upload
13 Here is a better oneliner that includes also the first line of each
16 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)}'
18 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
20 * memory leak in the syncher? It currently weighs 100M.
24 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
27 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
29 * lookup by epoch and by path and use this ability on the pause to never
30 again register a file twice that doesn't need it. Let's call it
33 * after the dirtymark is done: fill up recentfiles with fake (historic)
34 entries; fill up with individual corrections; algorithm maybe to be done
35 with bigfloat so that we can always place something in the middle
36 between two entries. Before we must switch to bigfloat we could try to
37 use Data::Float::nextup to get the.
39 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
42 * dirtymark now lives in Recentfile, needs to be used in rmirror.
44 * find out why the downloader died after a couple of hours without a net
45 connection. Write a test that survives the not-existence of the other
48 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
50 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
51 it makes less sense with tight coupling for secondary files. But asking
52 the server what the current epoch is might be cheaper on HTTP than on
53 rsync. (Needs to be evaluated)
55 * remove the 0.00 from the verbose overview in the Merged column in the
58 * write tests that expose the problems of the last few days: cascading
59 client/server roles, tight coupling for secondary RFs, deletes after
62 * Some day we might want to have policy options for the slave:
63 tight/loose/no coupling with upstream for secondary RFs. tight is what
64 we have now. loose would wait until a gap occurs that can be closed.
66 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
68 * revisit all $rfs->[$i+1] places if they now make sense still
70 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
72 * another bug is the fact that the mirror command deletes files before
73 it unhides the index file, thus confusing downstream slaves. We must not
74 delete before unhiding and must delete after unhiding. FIXED.
76 * new complication about the slave that is playing a server role.
77 Currently we mirror from newest to oldest with a hidden temporary file
78 as index. And when one file is finished, we unhide the index file.
79 Imagine the cascading server/slave is dead for a day. It then starts
80 mirroring again with the freshest thing and unhides the freshest index
81 file when it has worked through it. In that moment it exposes a time
82 hole. Because it now works on the second recentfile which is still
85 We currently do nothing special to converge after such a drop out. At
86 least not intentionally and robustly and thought through.
88 The algorithm we use to seed the next file needs quite a lot of more
89 robustness than it currently has. Something to do with looking at the
90 merged element of the next rf and when it has dropped off, we seed
91 immediately. And if it ramains dropped off, we seed again, of course.
93 Nope, looking from smaller to larger RFS we look at the merged element
94 of this RF and at the minmax/max element of the next RF. If that
95 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
98 And we need a public accessor seed and unseed or seeded. But not the mix
99 of public and private stuff that then is used behind the back.
101 And then the secondary* stuff must go.
103 And we must understand what the impact is on the DONE system. Can it go
104 unnoticed that there was a hole? And could the DONE system have decided
105 the hole is covered? This should be testable with three directories where
106 the middle stops working for a while. Done->merge is suspicious, we must
107 stop it from merging non-conflatable neighbors due to broken continuity.
111 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
113 * Slaven suggests to have the current epoch or the whole current
114 recentfile available from the HTTP server and take it away with
115 keepalive. This direction goes the granularity down to subseconds.
117 We might want to rewrite everything to factor out transport and allow
118 the whole thing to run via HTTP.
120 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
122 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
125 * smoker on k81 fetching from k75 to verify cascading works. See
126 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
128 * maybe the loop should wait for CHECKSUMS file after every upload. And
129 CPAN.pm needs to deal with timestamps in the future.
131 * do not forget the dirtymark!
133 Text: have a new flag on recentfiles with the meaning: if this
134 changes, you're required to run a full rsync over all the files. The
135 reason why we set it would probably be: some foul happened. we injected
136 files in arbitrary places or didn't inject them although they changed.
137 The content of the flag? Timestamp? The relation between the
138 recentfiles would have to be inheritance from the principal, because any
139 out of band changes would soon later propagate to the next recentfile.
141 By upping the flag often one can easily ruin the slaves.
143 last out of band change? dirtymark?
145 Anyway, this implies that we read a potentially existing recentfile
148 And it implies that we have an eventloop that keeps us busy in 2-3
149 cycles, one for current stuff (tight loop) and one for the recentfiles
150 (cascade when principal has changed), one for the old stuff after a
153 And it implies that the out-of-band change in any of the recentfiles
154 must have a lock on the principal file and there is the place to set the
157 * start a FAQ, especially quick start guide questions. Also to aid those
158 problematic areas where we have no good solution, like the "links"
161 * wish feedback when we are slow.
165 * Remove a few DEBUG statements.
167 * The multiple-rrr way of doing things needs a new option to rmirror,
168 like piecemeal or so. Not urgent because after the first pass through,
169 things run smoothely. It's only ugly during the first pass.
171 * I have the suspicion that the code is broken that decides if the
172 neighboring RF needs to be seeded. I fear when too much time has gone
173 between two calls (in our case more than one hour), it would not seed
174 the neighbor. Of course this will never be noticed, so we need a good
177 * local/localroot confusion: I currently pass both options but one must
180 * accounts for early birds on PAUSE rsync daemon.
182 * hardcoded 20 seconds
184 * who mirrors the index? DOING now.
186 * which CPAN mirrors offer rsync?
188 * visit all XXX, visit all _float places
190 * rename the pathdb stuff, it's too confusing. No idea how.
192 * rrr-inotify, backpan, rrr-register
194 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
196 * current bugs: the pathdb seems to get no reset, the seeding of the
197 secondaryttl stuff seems not to have an effect. Have helped myself with
198 a rand(10), need to fix this back. So not checked in. Does the rand
201 The rand thing helps. The secondaryttl stuff was in the wrong line,
204 The pathdb stuff was because I called either _pathdb or __pathdb on the
205 wrong object. FIXED now.
207 * It's not so beautiful if we never fetch the recentfiles that are not
208 the principal, even if this is correct behaviour. We really do not need
209 them after we have fetched the whole content.
211 OK, we want a switch for that: secondaryttl DONE
213 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
215 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
216 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
217 it lets 40 follow. FIXED
219 * See that the long running process really only updates the principal
220 file unless it has missed a timespan during which something happened. If
221 nothing happened, it must notice even when it misses the timespan. DONE
223 * we must throw away the pathdb when we have reached the end of Z. From
224 that moment we can have a very small pathdb because the only reason for
225 a pathdb is that we know to ignore old records in old files. We won't
226 need this pathdb again before the next full pass over the data is
227 necessary and then we will rebuild it as we go along. DONE
229 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
231 * I think, Done::register_one is doing wrong in that it does not
232 conflate neighboring pieces. The covered() method cannot do this because
233 it has no recent_events array at hand. But register_one has it and could
234 do it and for some reason misses to do it (sometimes).
236 This means that the three tests I just wrote can probably not survive
237 because they test with an already broken Done structure.
239 The art now is to detect how it happens, then to reproduce, then write a
242 So from the logfile this is what happens: we have a good interval with
243 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
244 top of it. Locally we now mirror F2 and open a new done interval for it.
245 Then we mirror F1 but this time with the timestamp T1b. And when we then
246 try to close the gap, we do not find T1 but instead something older. We
247 should gladly accept this older piece and this would fix this bug.
251 * bug to fix: when the 1h file changes while rmirror is running, we do
252 correctly sync the new files but never switch to the 6h file but rather
253 stay in a rather quick loop that fetches the 1h file again and again.
255 Is it possible that we initialize a new object? Or does
256 get_remote_recentfile_as_tempfile overwrite something in myself?
258 Want a new option: _runstatusfile => $file which frequently dumps the
259 state of all recentfiles to a file.
263 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
265 * Todo: now teach update to verify the timestamp is about to write
266 against the previous and use _increase_a_bit if it doesn't comply with
267 strict monotony. DONE
269 * The problem of rounding. So far perl's default precision was
270 sufficient. One day it won't be. FakeFloat has an easy job when it is
271 only reading and other machines have written correctly. But when we want
272 to write a floating point number that is a bit larger than the other
273 one, then we need our own idea of precision.
275 Slaven said: just append a "1". This might be going towards the end of
276 usability too quickly. I'd like something that actually uses the decimal
277 system. Well, appending a 1 also does this but...
279 E.g. we have 1.0. nextup on this architecture is starting with
280 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
281 taken the 1.0000000000000003 and the next user comes and the time tells
282 him 1.0 again. He has to beat my number without stepping over the
283 nextup. This is much less space than I had when I chose 1,2,3.
285 What is also irritating is that nextup is architecture dependent. The
286 128 bit guy must choose very long numbers to fit in between whereas the
287 other one with 16 bit uses larger steps. But then the algorithm is the
288 same for both, so that would be a nice thing.
290 I see two situation where we need this. One is when Time::HiRes returns
291 us a value that is <= the last entry in our recentfile. In this case
292 (let's call it the end-case) we must fill the region between that number
293 and the next higher native floating point number. The other is when we
294 inject an old file into an old recentfile (we would then also set a new
295 dirtymark). We find the integer value already taken and need a slightly
296 different one (let's call it the middle-case). The difference between
297 the two situations is that the next user will want to find something
298 higher than my number in the end-case and something lower than my number
301 So I suggest we give the function both a value and an upper bound and it
302 calculates us a primitive middle. The upper bound in the middle-case is
303 the next integer. The upper bound on the end-case is the nextup floating
304 point number. But the latter poses another problem: if we have occupied
305 the middle m between x and nextup(x), then the nextup(m) will probably
306 not be the same as nextup(x) because some rounding will take place
307 before the nextup is calculated and when the rounding reaches the
308 nextup(x), we will end up at nextup(nextup(x)).
310 So we really need to consider the nextup and the nextdown from there and
311 then the middle and that's the number we may approach asymptotically.
314 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
316 * consider deprecating the use of RECENT.recent as a symlink. It turns
317 out to need extra hoops with the rsync options and just isn't worth it.
318 Or maybe these extra hoops are needed anyway for the rest of the tree?
319 Nope, can't be the case because not all filesystems support symlinks.
321 But before doing the large step, I'll deprecate the call of
322 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
323 only introduced to resolve RECENT.recent and complicates the routine far
324 beyond what it deserves.
326 DONE. Won't deprecate RECENT.recent, just moved its handling to the
329 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
331 * I think it's a bug that the rsync_option links must be set to true in
332 order to support RECENT.recent and that nobody cares to set it
333 automatically. Similar for ignore_link_stat_errors.
335 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
337 * Todo: find all todos together and make a plan what is missing for a
340 - verifytree or something like that. fsck maybe.
342 - rersyncrecent, the script itself? What it do?
344 - a way to only mirror the recentfiles without mirroring the whole
345 remote system such that people can decide to mirror only partially see
346 also 2008-08-30. .shadow-xxx directory? this also needed for a
347 filesystem that is still incomplete and might need the mirrorfiles for
350 - long living objects that mirror again and again. Inject something
351 into ta, see how it goes over to tb.
353 - how do we continue filling up the DONE system when we use an object
354 for the second time? "fully covered" and "uptodate" or new terminology.
356 - overview called on the wrong file should be understandable
358 - the meta data field that must change when we fake something up so that
359 the downstream people know they have to re-fetch everything.
361 - how tolerant are we against missing files upstream? how do we keep
362 track? there are legitimate cases where we did read upstream index right
363 before a file got deleted there and then find that file as new and want
364 it. There are other cases that are not self healing and must be tracked
367 - how, exactly, do we have to deal with deletes? With rsync errors?
369 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
370 authors) failed: No such file or directory (2)
372 The file above is a delete in 1h and a new in file 1M and the
373 delete in the locally running rmirror did not get propagated to the 1M
374 object. Bug. And the consequence is a standstill.
376 It seems that a slave that works with a file below the principal needs
377 to merge things all the way up to get rid of later deletes. Or keep
378 track of all deletes and skip them later. So we need a trackdeletes.pm
379 similar to the done.pm?
381 see also 2008-08-20 about spurious deletes that really have no add
382 counterpart and yet they are not wrong.
384 - consider the effect when resyncing the recentfile takes longer than
385 the time per loop. Then we never rsync any file. We need to diagnose
386 that and force an increase of that loop time. But when we later are fast
387 enough again because the net has recovered, then we need to switch back
388 to original parameters. ERm, no, it's enough to keep syncing at least
389 one file before refetching an index file.
391 - remember to verify that no temp files are left lying around and the
394 - status file for not long running jobs that want to track upstream with
397 - revisit all XXX _float areas and study Sub::Exporter DONE
399 - persistent DB even though we just said we do not need it. Just for
400 extended capabilities and time savings when, for example, upstream
401 announces a reset and we get new recentfiles and could then limit
402 ourselves to a subset of files (those that have a changed epoch) in a
403 first pass and would only then do the loop to verify the rest. Or
406 * Todo: aggregate files should know their feed and finding the principal
407 should be done stepwise. (?)
409 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
410 left locks around. DONE
412 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
414 * maybe extend the _overview so that it always says if and where the
415 last file is in the next file and where the next event in the next rf
416 would lie. No, don't like this anymore. REJECT
418 * take the two new redundant tests out again, only the third must
421 * Todo: add a sanity check if the merged structure is really pointing to
422 a different rf and that this different rf is larger. DONE
424 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
426 * now test, if they are overlapping. And test if there is a file in the
427 next rf that would fit into this rf's interval.
429 1h 1222324012.8474 1222322541.7963 0.4086
430 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
431 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
432 1W 1222313218.3626 1221708477.5829 167.9835
434 I suspect that somebody writes a merged timestamp without having merged
435 and then somebody else relies on it.
437 If aggregate is running, the intervals must not be extravagated, if it
438 is not running, there must not be bounds, the total number of events in
439 the system must be counted and must be controlled throughout the tests.
440 That the test required the additional update was probably nonsense,
441 because aggregate can cut pieces too. FIXED & DONE
443 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
445 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
448 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
450 * Most apparent bug at the moment is that the recentfiles are fetched
451 too often. Only the principal should be fetched and if it has not
452 changed, the others should not be refetched. ATM I must admit that I'm
453 happy that we refetch more often than needed because I can more easily
454 fix bugs while the thing is running.
456 * Let's say, 1220474966.19501 is a timestamp of a file that is already
457 done but the done system does not know about it. The reason for the
458 failure is not known and we never reach the status uptodate because of
459 this. We must get over it.
461 Later it turns out that the origin server had a bug somewhere.
462 1220474966.19042 came after 1220474966.19501. Or better: it was in the
463 array of the recentfile one position above. The bug was my own.
465 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
467 * There is the race condition where the server does a delete and the
468 slave does not yet know and then tries to download it because he sees
469 the new. So for this time window we must be more tolerant against
470 failure. If we cannot download a file, we should just skip it and should
471 not retry immediately. The whole system should discover the lost thing
472 later. Keeping track with the DONE system should really be a no brainer.
474 But there is something more: the whole filesystem is a database and the
475 recentfiles are one possible representation of it. It's a pretty useful
476 representation I think that's why I have implemented something around
477 it. But for strictly local operation it has little value. For local
478 operation we would much rather have a database. So we would enter every
479 recentfile reading and every rsync operation and for every file the last
480 state change and what it leads to. Then we would always ignore older
481 records without the efforts involved with recentfiles.
483 The database would have: path,recentepoch,rsyncedon,deletedon
485 Oh well, not yet clear where this leads to.
487 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
489 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
490 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
492 * statusfile, maybe only on demand, alone to have a sharp debugging
493 tool. It is locked and all recentfiles dump themselves into it and we
494 can build a viewer that lets us know where we stand and what's inside.
496 * remember: only the principal recentfile needs expiration, all others
497 shall be expired by principal if it discovers that something has move
500 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
502 * Always check if we stringify to a higher value than in the entry
505 * And in covered make an additional check if we would be able to see a
506 numerical difference between the two numbers and if we can't then switch
507 to a different, more expensive algorithm. Do not want to be caught by
508 floating surprises. DONE
510 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
512 * caching has several aspects here: we can cache the interval of the
513 recentfile which only will change when the mtime of the file changes. We
514 must re-mirror the recentfile when its ttl has expired. Does have_read
515 tell you anything? It counts nothing at all. Only the mtime is
516 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
517 thing the have_mirrored because it is unrelated to the mtime.
519 * Robustness of floating point calculations! I always thought that the
520 string calculated by the origin server for the floating representation
521 of the epoch time is just a string. When we convert it to a number and
522 later back to a string, the other computer might come to a different
523 conclusion. This must not happen, we want to preserve it under any
524 circumstances. I will have to write tests with overlong sequences that
525 get lost in arithmetic and must see if all still works well. DONE
527 But one fragile point remains: if one host considers a>b and the other
528 one considers them == but no eq. To prevent this, we must probably do
529 some extra homework. DONE
531 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
533 * the concept of tracking DONE needs an object per recentfile that has
534 something like these methods:
536 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
538 covered() register() covered()
540 The unclear thing is how we translate points in time into intervals. We
541 could pass a reference to the current recent_events array when running
542 we_have(xxx) and let the DONE object iterate over it such that it only
543 has to store a list of intervals that can melt into each other. Ah, even
544 passing the list together with a list of indexes seems feasiable.
546 Or maybe ask for the inverted list?
548 Whenever the complete array is covered by the interval we say we are
549 fully covered and if the recentfile is not expired, we are uptodate.
551 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
553 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
555 * need a way to "return" the next entry after the end of a list. When
556 the caller says "before" or "after" we would like to know if he could
557 cover that interval/threshold or not because this influences the effect
558 of a newer timestamp of that recentfile. DONE with $opt{info}.
560 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
562 * one of the next things to tackle: the equivalent of csync2 -TIXU.
564 loop implies tixu (?). Nope, something like --statefile decides. Per
567 T test, I init, X including removals, U nodirtymark
569 So we have no concept of dirtymarks, we only trust that since we are
570 running we have observed everything steadily. But people will not let
571 this program run forever so we must consider both startup penalty and
572 book keeping for later runs. We keep this for later. For now we write a
573 long running mirror that merges several intervals.
575 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
577 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
580 * rersyncrecent, the script: default to one week. The name of the switch
581 is --after. Other switches? --loop!
583 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
585 * need a switch --skip-deletes (?)
587 * need a switch --enduser that tells us that the whole tempfile
588 discipline is not needed when there is no downstream user. (?)
590 Without this switch we cannot have a reasonable recent.pl that just
591 displays the recent additions. Either we accept to download everything.
592 Or we download temporary files without the typical rsync protocol
595 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
596 File::Temp::tempdir, this might be a win.
598 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
600 * apropos missing: we have no push, we never know the downstream
601 servers. People who know their downstream hosts and want to ascertain
602 something will want additional methods we have never thought about, like
603 update or delete a certain file.
605 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
607 * tempted to refactor rmirror into resolve_symlink, localize, etc.
608 Curious if rsync_options=links equal 0 vs. 1 will make the expected
611 * rsync options: it's a bit of a pain that we usually need several rsync
612 options, like compress, links, times, checksum and that there is no
613 reasonable default except the original rsync default. I think wee can
614 safely assume that the rsync options are shared between all recentfile
615 instances within one recent tree.
617 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
619 * deletes: if a delete follows an add quickly enough it may happen that
620 a downstream mirror did not see the add at all! It seems this needs to
621 be mentioned somewhere. The point here is that even if the downstream is
622 never missing the principal timeframe it may encounter a "delete" that
623 has no complimentary "add" anywhere.
625 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
627 * I suspect the treat of metadata is incorrect during read or something.
628 The bug that I am watching is that between 06:08 and 06:09 the 6h file
629 contained more than 6 hours worth of data. At 06:08 we merged into the
630 1d file. We need to take snapshots of the 6h file over the course of an
631 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
634 Much worse: watching the 1h file: right at the moment (at 06:35) it
635 covers 1218867584-1219120397 which is 70 hours.
637 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
638 08:19:44 2008, that is when I checked out last time, so it seems to be
639 aggregating and never truncating?
641 No, correct is: it is never truncating; but wrong is: it is aggregating.
642 It does receive a lot of events from time to time from a larger file.
643 Somehow a large file gets merged into the small one and because the
644 "meta/merged" attribute is missing, nobody is paying attention. I
645 believe that I can fix this by making sure that metadata are honoured
646 during read. DONE and test adjusted.
648 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
650 * grand renaming plan
652 remotebase => remoteroot to fit well with localroot DONE
653 local_path() => localroot seems to me should already work DONE
654 recentfile_basename => rfilename no need to stress it has no slash DONE
656 filenameroot??? Doesn't seem too bad to me today. Maybe something like
657 kern? It would anyway need a deprecation cycle because it is an
658 important constructor.
660 * I like the portability that Data::Serializer brings us but the price
661 is that some day we might find out that it is slowing us a bit. We'll
664 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
666 * should we not enter the interval of the principal (or the interval of
667 the merging file?) in every aggregated/merged file?
669 * we should aim at a first release and give up on thinking about
670 sanitizing stuff and zloop. Let's just admit that a full traditional
671 rsync is the only available sanitizer ATM. Otherwise it's complicated
672 stuff: sanitizing on the origin server, sanitizing on the slaves,
673 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
674 get the basics out before this becomes a major cause for mess.
676 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
678 * On OSes not supporting symlinks we expect that RECENT.recent contains
679 the contents of the principal recentfile. Actually this is identical on
680 systems supporting symlinks. Simple, what follows from that is that we
681 need to keep the serializer in the metadata because we cannot read it
682 from the filename, doesn't it? Of course not. It's a chicken and egg
683 problem. This leaves us with the problem to actually parse the
684 serialized data to find out in which format it is. So who can do the 4
685 or 5 magics we wanted to support? File::LibMagic?
687 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
689 * remotebase and recentfile_basename are ugly names. Now that we need a
690 word for the shortest/principal/driving recentfile too we should do
693 localroot is good. rfile is good. local_path() is bad, local_path($path)
694 is medium, filenameroot() is bad, remotebase is bad, recentfile is
697 Up to now remotebase was the string that described the remote root
698 directory in rsync notation, like pause.perl.org::authors. And
699 recentfile_basename was "RECENT-1h.yaml".
701 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
703 * The test that was added in today's checkin is a good start for a test
704 of rmirror. We should have more methods in Recent.pm: verify,
705 addmissingfiles. We should verify the current tree, then rmirror it and
706 then verifytree the copy. We could then add some arbitrary file and let
707 it be discovered by addmissingfiles, then rmirror again and then
708 verifytree the copy again.
710 Then we could start stealing from csync2 sqlite database [no port to
711 OSX!] and fill a local DB. And methods to compare the database with the
712 recentfiles. Our strength is that in principle we could maintain state
713 with a single float. We have synced up to 1234567890.123456. If the Z
714 file does not add new files all we have to do is mirror the new ones and
717 This makes it clear that we should extend current protocol and declare
718 that we cheat when we add files too late, just to help the other end
719 keeping track. Ah yes, that's what was meant when zloop was mentioned
722 Maybe need to revisit File::Mirror to help me with this task.
724 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
726 * There must be an allow-me-to-truncate flag in every recentfile.
727 Without it one could construct a sequence of updates winning the locking
728 battle against the aggregator. Only if an aggregator has managed to
729 merge data over to the next level, truncating can be allowed. DONE with
732 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
734 * We should probably guarantee that no duplicates enter the aggregator
737 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
739 * To get merge operation faster would need a good benchmark test. What
740 02 spits out isn't reliable enough and is dominated by many other
743 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
744 Author: Andreas J. Koenig <k@k75.(none)>
745 Date: Sat Aug 2 07:58:04 2008 +0200
749 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
750 Author: Andreas J. Koenig <k@k75.(none)>
751 Date: Sat Aug 2 11:40:29 2008 +0200
753 there was a lot of trying but the effect is hardly measurable with
756 * overhead of connecting seems high. When setting
757 max_files_per_connection to 1 we see that.
759 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
761 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
763 25138.2215600014/3600 = 6.98283932222261
765 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
766 be a bug in the aggregator. FIXED
768 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
770 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
771 we have it, can we mirror it, mirror it!
773 I fear this needs a new class which might be called
774 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
775 and treat them as an entity. I realize that a single recentfile may be
776 sufficient for certain tasks and that it is handy for the low level
777 programmer but it is not nice to use. If there is a delete in the 1h
778 file then the 6h file still contains it. Seekers of the best information
779 need to combine at least some of the recentfiles most of the time.
781 There is the place for the Z loop!
783 But the combination is something to collect in a database, isn't it. Did
784 csync2 just harrumph?
786 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
788 * it just occurred to me that hosts in the same mirroring pool could
789 help out each other even without rewriting the recentfile. Just fetch
790 the stuff to mirror from several places, bingo. But that's something
791 that should rather live in a separate package or in rsync directly.
793 * cronjobs are unsuited because with ntp they would all come at the full
794 minute and disturb each other. Besides that I'd hate to have a backbone
795 with more than a few seconds latency.
797 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
799 * a second rsync server with access control for PAUSE. Port? 873 is the
800 standard port, let's take 8873.
802 * if there were a filesystem based on this, it would have a slow access
803 to inexistent files. It would probably provide wrong readdir (only based
804 on current content) or also a slow one (based on a recentfile written
805 after the call). But it would provide fast access to existing files. Or
806 one would deliberately allow slightly blurred answers based on some
807 sqlite reflection of the recentfiles.
809 * todo: write a variant of mirror() that combines two or more
810 recentfiles and treats them like one
812 * todo: signal handler to remove the tempfile
814 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
816 * now that we have the symlink I forgot how it should be used in
819 * the z loop: add missing files to Z file. Just append them (instead of
820 prepending). So one guy prepends something from the Y file from time to
821 time and another guy appends something rather frequently. Collecting
822 pond. When Y merges into Z, things get epoch and the collecting pond
823 gets smaller. What exactly are "missing files"?
825 take note of current epoch of the alpha file, let's call it the
828 find all files on disk
830 remove all files registered in the recentworld up to recent-ts
832 remove all files that have been deleted after recent-ts according to
835 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
837 * rersyncrecent might be a cronjob with a (locked) state file which
838 contains things like after and maybe last z sync or such?
840 rrr-mirror might be an alternative name but how would we justify the
841 three Rs when there is no Re-Rsync-Recent?
843 With the --loop parameter it is an endless loop, without it is no loop.
844 At least this is simple.
846 * todo: new accssor z-interval specifies how often the Z file is updated
847 against the filesystem. We probably want no epoch stamp on these
848 entries. And we want to be able to filter the entries (e.g. no
849 by-modules and by-category tree)
851 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
853 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
854 into Z. We do not need the other files filled up with filesystem
857 * need interface to query for a file in order to NOT call update on
858 PAUSE a second time within a short time.
860 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
862 * recommended update interval? Makes no sense, is different for
869 change-log-default-name: "Todo"