1 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
4 it makes less sense with tight coupling for secondary files. But asking
5 the server what the current epoch is might be cheaper on HTTP than on
6 rsync. (Needs to be evaluated)
8 * remove the 0.00 from the verbose overview in the Merged column in the
11 * write tests that expose the problems of the last few days: cascading
12 client/server roles, tight coupling for secondary RFs, deletes after
15 * Some day we might want to have policy options for the slave:
16 tight/loose/no coupling with upstream for secondary RFs. tight is what
17 we have now. loose would wait until a gap occurs that can be closed.
19 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
21 * revisit all $rfs->[$i+1] places if they now make sense still
23 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
25 * another bug is the fact that the mirror command deletes files before
26 it unhides the index file, thus confusing downstream slaves. We must not
27 delete before unhiding and must delete after unhiding. FIXED.
29 * new complication about the slave that is playing a server role.
30 Currently we mirror from newest to oldest with a hidden temporary file
31 as index. And when one file is finished, we unhide the index file.
32 Imagine the cascading server/slave is dead for a day. It then starts
33 mirroring again with the freshest thing and unhides the freshest index
34 file when it has worked through it. In that moment it exposes a time
35 hole. Because it now works on the second recentfile which is still
38 We currently do nothing special to converge after such a drop out. At
39 least not intentionally and robustly and thought through.
41 The algorithm we use to seed the next file needs quite a lot of more
42 robustness than it currently has. Something to do with looking at the
43 merged element of the next rf and when it has dropped off, we seed
44 immediately. And if it ramains dropped off, we seed again, of course.
46 Nope, looking from smaller to larger RFS we look at the merged element
47 of this RF and at the minmax/max element of the next RF. If that
48 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
51 And we need a public accessor seed and unseed or seeded. But not the mix
52 of public and private stuff that then is used behind the back.
54 And then the secondary* stuff must go.
56 And we must understand what the impact is on the DONE system. Can it go
57 unnoticed that there was a hole? And could the DONE system have decided
58 the hole is covered? This should be testable with three directories where
59 the middle stops working for a while. Done->merge is suspicious, we must
60 stop it from merging non-conflatable neighbors due to broken continuity.
64 * dirtymark now lives in Recentfile, needs to be used in rmirror.
66 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
68 * Slaven suggests to have the current epoch or the whole current
69 recentfile available from the HTTP server and take it away with
70 keepalive. This direction goes the granularity down to subseconds.
72 We might want to rewrite everything to factor out transport and allow
73 the whole thing to run via HTTP.
75 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
77 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
80 * smoker on k81 fetching from k75 to verify cascading works. See
81 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
83 * maybe the loop should wait for CHECKSUMS file after every upload. And
84 CPAN.pm needs to deal with timestamps in the future.
86 * do not forget the dirtymark!
88 Text: have a new flag on recentfiles with the meaning: if this
89 changes, you're required to run a full rsync over all the files. The
90 reason why we set it would probably be: some foul happened. we injected
91 files in arbitrary places or didn't inject them although they changed.
92 The content of the flag? Timestamp? The relation between the
93 recentfiles would have to be inheritance from the principal, because any
94 out of band changes would soon later propagate to the next recentfile.
96 By upping the flag often one can easily ruin the slaves.
98 last out of band change? dirtymark?
100 Anyway, this implies that we read a potentially existing recentfile
103 And it implies that we have an eventloop that keeps us busy in 2-3
104 cycles, one for current stuff (tight loop) and one for the recentfiles
105 (cascade when principal has changed), one for the old stuff after a
108 And it implies that the out-of-band change in any of the recentfiles
109 must have a lock on the principal file and there is the place to set the
112 * after the dirtymark is done: fill up recentfiles with fake (historic)
113 entries; fill up with individual corrections; algorithm maybe to be done
114 with bigfloat so that we can always place something in the middle
115 between two entries. Before we must switch to bigfloat we could try to
116 use Data::Float::nextup to get the.
118 * lookup by epoch and by path and use this ability on the pause to never
119 again register a file twice that doesn't need it.
121 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
124 * start a FAQ, especially quick start guide questions. Also to aid those
125 problematic areas where we have no good solution, like the "links"
128 * wish feedback when we are slow.
132 * Remove a few DEBUG statements.
134 * The multiple-rrr way of doing things needs a new option to rmirror,
135 like piecemeal or so. Not urgent because after the first pass through,
136 things run smoothely. It's only ugly during the first pass.
138 * I have the suspicion that the code is broken that decides if the
139 neighboring RF needs to be seeded. I fear when too much time has gone
140 between two calls (in our case more than one hour), it would not seed
141 the neighbor. Of course this will never be noticed, so we need a good
144 * local/localroot confusion: I currently pass both options but one must
147 * accounts for early birds on PAUSE rsync daemon.
149 * hardcoded 20 seconds
151 * who mirrors the index? DOING now.
153 * which CPAN mirrors offer rsync?
155 * visit all XXX, visit all _float places
157 * rename the pathdb stuff, it's too confusing. No idea how.
159 * rrr-inotify, backpan, rrr-register
161 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
163 * current bugs: the pathdb seems to get no reset, the seeding of the
164 secondaryttl stuff seems not to have an effect. Have helped myself with
165 a rand(10), need to fix this back. So not checked in. Does the rand
168 The rand thing helps. The secondaryttl stuff was in the wrong line,
171 The pathdb stuff was because I called either _pathdb or __pathdb on the
172 wrong object. FIXED now.
174 * It's not so beautiful if we never fetch the recentfiles that are not
175 the principal, even if this is correct behaviour. We really do not need
176 them after we have fetched the whole content.
178 OK, we want a switch for that: secondaryttl DONE
180 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
182 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
183 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
184 it lets 40 follow. FIXED
186 * See that the long running process really only updates the principal
187 file unless it has missed a timespan during which something happened. If
188 nothing happened, it must notice even when it misses the timespan. DONE
190 * we must throw away the pathdb when we have reached the end of Z. From
191 that moment we can have a very small pathdb because the only reason for
192 a pathdb is that we know to ignore old records in old files. We won't
193 need this pathdb again before the next full pass over the data is
194 necessary and then we will rebuild it as we go along. DONE
196 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
198 * I think, Done::register_one is doing wrong in that it does not
199 conflate neighboring pieces. The covered() method cannot do this because
200 it has no recent_events array at hand. But register_one has it and could
201 do it and for some reason misses to do it (sometimes).
203 This means that the three tests I just wrote can probably not survive
204 because they test with an already broken Done structure.
206 The art now is to detect how it happens, then to reproduce, then write a
209 So from the logfile this is what happens: we have a good interval with
210 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
211 top of it. Locally we now mirror F2 and open a new done interval for it.
212 Then we mirror F1 but this time with the timestamp T1b. And when we then
213 try to close the gap, we do not find T1 but instead something older. We
214 should gladly accept this older piece and this would fix this bug.
218 * bug to fix: when the 1h file changes while rmirror is running, we do
219 correctly sync the new files but never switch to the 6h file but rather
220 stay in a rather quick loop that fetches the 1h file again and again.
222 Is it possible that we initialize a new object? Or does
223 get_remote_recentfile_as_tempfile overwrite something in myself?
225 Want a new option: _runstatusfile => $file which frequently dumps the
226 state of all recentfiles to a file.
230 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
232 * Todo: now teach update to verify the timestamp is about to write
233 against the previous and use _increase_a_bit if it doesn't comply with
234 strict monotony. DONE
236 * The problem of rounding. So far perl's default precision was
237 sufficient. One day it won't be. FakeFloat has an easy job when it is
238 only reading and other machines have written correctly. But when we want
239 to write a floating point number that is a bit larger than the other
240 one, then we need our own idea of precision.
242 Slaven said: just append a "1". This might be going towards the end of
243 usability too quickly. I'd like something that actually uses the decimal
244 system. Well, appending a 1 also does this but...
246 E.g. we have 1.0. nextup on this architecture is starting with
247 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
248 taken the 1.0000000000000003 and the next user comes and the time tells
249 him 1.0 again. He has to beat my number without stepping over the
250 nextup. This is much less space than I had when I chose 1,2,3.
252 What is also irritating is that nextup is architecture dependent. The
253 128 bit guy must choose very long numbers to fit in between whereas the
254 other one with 16 bit uses larger steps. But then the algorithm is the
255 same for both, so that would be a nice thing.
257 I see two situation where we need this. One is when Time::HiRes returns
258 us a value that is <= the last entry in our recentfile. In this case
259 (let's call it the end-case) we must fill the region between that number
260 and the next higher native floating point number. The other is when we
261 inject an old file into an old recentfile (we would then also set a new
262 dirtymark). We find the integer value already taken and need a slightly
263 different one (let's call it the middle-case). The difference between
264 the two situations is that the next user will want to find something
265 higher than my number in the end-case and something lower than my number
268 So I suggest we give the function both a value and an upper bound and it
269 calculates us a primitive middle. The upper bound in the middle-case is
270 the next integer. The upper bound on the end-case is the nextup floating
271 point number. But the latter poses another problem: if we have occupied
272 the middle m between x and nextup(x), then the nextup(m) will probably
273 not be the same as nextup(x) because some rounding will take place
274 before the nextup is calculated and when the rounding reaches the
275 nextup(x), we will end up at nextup(nextup(x)).
277 So we really need to consider the nextup and the nextdown from there and
278 then the middle and that's the number we may approach asymptotically.
281 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
283 * consider deprecating the use of RECENT.recent as a symlink. It turns
284 out to need extra hoops with the rsync options and just isn't worth it.
285 Or maybe these extra hoops are needed anyway for the rest of the tree?
286 Nope, can't be the case because not all filesystems support symlinks.
288 But before doing the large step, I'll deprecate the call of
289 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
290 only introduced to resolve RECENT.recent and complicates the routine far
291 beyond what it deserves.
293 DONE. Won't deprecate RECENT.recent, just moved its handling to the
296 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
298 * I think it's a bug that the rsync_option links must be set to true in
299 order to support RECENT.recent and that nobody cares to set it
300 automatically. Similar for ignore_link_stat_errors.
302 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
304 * Todo: find all todos together and make a plan what is missing for a
307 - verifytree or something like that. fsck maybe.
309 - rersyncrecent, the script itself? What it do?
311 - a way to only mirror the recentfiles without mirroring the whole
312 remote system such that people can decide to mirror only partially see
313 also 2008-08-30. .shadow-xxx directory? this also needed for a
314 filesystem that is still incomplete and might need the mirrorfiles for
317 - long living objects that mirror again and again. Inject something
318 into ta, see how it goes over to tb.
320 - how do we continue filling up the DONE system when we use an object
321 for the second time? "fully covered" and "uptodate" or new terminology.
323 - overview called on the wrong file should be understandable
325 - the meta data field that must change when we fake something up so that
326 the downstream people know they have to re-fetch everything.
328 - how tolerant are we against missing files upstream? how do we keep
329 track? there are legitimate cases where we did read upstream index right
330 before a file got deleted there and then find that file as new and want
331 it. There are other cases that are not self healing and must be tracked
334 - how, exactly, do we have to deal with deletes? With rsync errors?
336 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
337 authors) failed: No such file or directory (2)
339 The file above is a delete in 1h and a new in file 1M and the
340 delete in the locally running rmirror did not get propagated to the 1M
341 object. Bug. And the consequence is a standstill.
343 It seems that a slave that works with a file below the principal needs
344 to merge things all the way up to get rid of later deletes. Or keep
345 track of all deletes and skip them later. So we need a trackdeletes.pm
346 similar to the done.pm?
348 see also 2008-08-20 about spurious deletes that really have no add
349 counterpart and yet they are not wrong.
351 - consider the effect when resyncing the recentfile takes longer than
352 the time per loop. Then we never rsync any file. We need to diagnose
353 that and force an increase of that loop time. But when we later are fast
354 enough again because the net has recovered, then we need to switch back
355 to original parameters. ERm, no, it's enough to keep syncing at least
356 one file before refetching an index file.
358 - remember to verify that no temp files are left lying around and the
361 - status file for not long running jobs that want to track upstream with
364 - revisit all XXX _float areas and study Sub::Exporter DONE
366 - persistent DB even though we just said we do not need it. Just for
367 extended capabilities and time savings when, for example, upstream
368 announces a reset and we get new recentfiles and could then limit
369 ourselves to a subset of files (those that have a changed epoch) in a
370 first pass and would only then do the loop to verify the rest. Or
373 * Todo: aggregate files should know their feed and finding the principal
374 should be done stepwise. (?)
376 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
377 left locks around. DONE
379 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
381 * maybe extend the _overview so that it always says if and where the
382 last file is in the next file and where the next event in the next rf
383 would lie. No, don't like this anymore. REJECT
385 * take the two new redundant tests out again, only the third must
388 * Todo: add a sanity check if the merged structure is really pointing to
389 a different rf and that this different rf is larger. DONE
391 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
393 * now test, if they are overlapping. And test if there is a file in the
394 next rf that would fit into this rf's interval.
396 1h 1222324012.8474 1222322541.7963 0.4086
397 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
398 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
399 1W 1222313218.3626 1221708477.5829 167.9835
401 I suspect that somebody writes a merged timestamp without having merged
402 and then somebody else relies on it.
404 If aggregate is running, the intervals must not be extravagated, if it
405 is not running, there must not be bounds, the total number of events in
406 the system must be counted and must be controlled throughout the tests.
407 That the test required the additional update was probably nonsense,
408 because aggregate can cut pieces too. FIXED & DONE
410 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
412 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
415 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
417 * Most apparent bug at the moment is that the recentfiles are fetched
418 too often. Only the principal should be fetched and if it has not
419 changed, the others should not be refetched. ATM I must admit that I'm
420 happy that we refetch more often than needed because I can more easily
421 fix bugs while the thing is running.
423 * Let's say, 1220474966.19501 is a timestamp of a file that is already
424 done but the done system does not know about it. The reason for the
425 failure is not known and we never reach the status uptodate because of
426 this. We must get over it.
428 Later it turns out that the origin server had a bug somewhere.
429 1220474966.19042 came after 1220474966.19501. Or better: it was in the
430 array of the recentfile one position above. The bug was my own.
432 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
434 * There is the race condition where the server does a delete and the
435 slave does not yet know and then tries to download it because he sees
436 the new. So for this time window we must be more tolerant against
437 failure. If we cannot download a file, we should just skip it and should
438 not retry immediately. The whole system should discover the lost thing
439 later. Keeping track with the DONE system should really be a no brainer.
441 But there is something more: the whole filesystem is a database and the
442 recentfiles are one possible representation of it. It's a pretty useful
443 representation I think that's why I have implemented something around
444 it. But for strictly local operation it has little value. For local
445 operation we would much rather have a database. So we would enter every
446 recentfile reading and every rsync operation and for every file the last
447 state change and what it leads to. Then we would always ignore older
448 records without the efforts involved with recentfiles.
450 The database would have: path,recentepoch,rsyncedon,deletedon
452 Oh well, not yet clear where this leads to.
454 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
456 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
457 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
459 * statusfile, maybe only on demand, alone to have a sharp debugging
460 tool. It is locked and all recentfiles dump themselves into it and we
461 can build a viewer that lets us know where we stand and what's inside.
463 * remember: only the principal recentfile needs expiration, all others
464 shall be expired by principal if it discovers that something has move
467 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
469 * Always check if we stringify to a higher value than in the entry
472 * And in covered make an additional check if we would be able to see a
473 numerical difference between the two numbers and if we can't then switch
474 to a different, more expensive algorithm. Do not want to be caught by
475 floating surprises. DONE
477 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
479 * caching has several aspects here: we can cache the interval of the
480 recentfile which only will change when the mtime of the file changes. We
481 must re-mirror the recentfile when its ttl has expired. Does have_read
482 tell you anything? It counts nothing at all. Only the mtime is
483 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
484 thing the have_mirrored because it is unrelated to the mtime.
486 * Robustness of floating point calculations! I always thought that the
487 string calculated by the origin server for the floating representation
488 of the epoch time is just a string. When we convert it to a number and
489 later back to a string, the other computer might come to a different
490 conclusion. This must not happen, we want to preserve it under any
491 circumstances. I will have to write tests with overlong sequences that
492 get lost in arithmetic and must see if all still works well. DONE
494 But one fragile point remains: if one host considers a>b and the other
495 one considers them == but no eq. To prevent this, we must probably do
496 some extra homework. DONE
498 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
500 * the concept of tracking DONE needs an object per recentfile that has
501 something like these methods:
503 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
505 covered() register() covered()
507 The unclear thing is how we translate points in time into intervals. We
508 could pass a reference to the current recent_events array when running
509 we_have(xxx) and let the DONE object iterate over it such that it only
510 has to store a list of intervals that can melt into each other. Ah, even
511 passing the list together with a list of indexes seems feasiable.
513 Or maybe ask for the inverted list?
515 Whenever the complete array is covered by the interval we say we are
516 fully covered and if the recentfile is not expired, we are uptodate.
518 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
520 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
522 * need a way to "return" the next entry after the end of a list. When
523 the caller says "before" or "after" we would like to know if he could
524 cover that interval/threshold or not because this influences the effect
525 of a newer timestamp of that recentfile. DONE with $opt{info}.
527 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
529 * one of the next things to tackle: the equivalent of csync2 -TIXU.
531 loop implies tixu (?). Nope, something like --statefile decides. Per
534 T test, I init, X including removals, U nodirtymark
536 So we have no concept of dirtymarks, we only trust that since we are
537 running we have observed everything steadily. But people will not let
538 this program run forever so we must consider both startup penalty and
539 book keeping for later runs. We keep this for later. For now we write a
540 long running mirror that merges several intervals.
542 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
544 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
547 * rersyncrecent, the script: default to one week. The name of the switch
548 is --after. Other switches? --loop!
550 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
552 * need a switch --skip-deletes (?)
554 * need a switch --enduser that tells us that the whole tempfile
555 discipline is not needed when there is no downstream user. (?)
557 Without this switch we cannot have a reasonable recent.pl that just
558 displays the recent additions. Either we accept to download everything.
559 Or we download temporary files without the typical rsync protocol
562 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
563 File::Temp::tempdir, this might be a win.
565 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
567 * apropos missing: we have no push, we never know the downstream
568 servers. People who know their downstream hosts and want to ascertain
569 something will want additional methods we have never thought about, like
570 update or delete a certain file.
572 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
574 * tempted to refactor rmirror into resolve_symlink, localize, etc.
575 Curious if rsync_options=links equal 0 vs. 1 will make the expected
578 * rsync options: it's a bit of a pain that we usually need several rsync
579 options, like compress, links, times, checksum and that there is no
580 reasonable default except the original rsync default. I think wee can
581 safely assume that the rsync options are shared between all recentfile
582 instances within one recent tree.
584 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
586 * deletes: if a delete follows an add quickly enough it may happen that
587 a downstream mirror did not see the add at all! It seems this needs to
588 be mentioned somewhere. The point here is that even if the downstream is
589 never missing the principal timeframe it may encounter a "delete" that
590 has no complimentary "add" anywhere.
592 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
594 * I suspect the treat of metadata is incorrect during read or something.
595 The bug that I am watching is that between 06:08 and 06:09 the 6h file
596 contained more than 6 hours worth of data. At 06:08 we merged into the
597 1d file. We need to take snapshots of the 6h file over the course of an
598 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
601 Much worse: watching the 1h file: right at the moment (at 06:35) it
602 covers 1218867584-1219120397 which is 70 hours.
604 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
605 08:19:44 2008, that is when I checked out last time, so it seems to be
606 aggregating and never truncating?
608 No, correct is: it is never truncating; but wrong is: it is aggregating.
609 It does receive a lot of events from time to time from a larger file.
610 Somehow a large file gets merged into the small one and because the
611 "meta/merged" attribute is missing, nobody is paying attention. I
612 believe that I can fix this by making sure that metadata are honoured
613 during read. DONE and test adjusted.
615 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
617 * grand renaming plan
619 remotebase => remoteroot to fit well with localroot DONE
620 local_path() => localroot seems to me should already work DONE
621 recentfile_basename => rfilename no need to stress it has no slash DONE
623 filenameroot??? Doesn't seem too bad to me today. Maybe something like
624 kern? It would anyway need a deprecation cycle because it is an
625 important constructor.
627 * I like the portability that Data::Serializer brings us but the price
628 is that some day we might find out that it is slowing us a bit. We'll
631 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
633 * should we not enter the interval of the principal (or the interval of
634 the merging file?) in every aggregated/merged file?
636 * we should aim at a first release and give up on thinking about
637 sanitizing stuff and zloop. Let's just admit that a full traditional
638 rsync is the only available sanitizer ATM. Otherwise it's complicated
639 stuff: sanitizing on the origin server, sanitizing on the slaves,
640 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
641 get the basics out before this becomes a major cause for mess.
643 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
645 * On OSes not supporting symlinks we expect that RECENT.recent contains
646 the contents of the principal recentfile. Actually this is identical on
647 systems supporting symlinks. Simple, what follows from that is that we
648 need to keep the serializer in the metadata because we cannot read it
649 from the filename, doesn't it? Of course not. It's a chicken and egg
650 problem. This leaves us with the problem to actually parse the
651 serialized data to find out in which format it is. So who can do the 4
652 or 5 magics we wanted to support? File::LibMagic?
654 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
656 * remotebase and recentfile_basename are ugly names. Now that we need a
657 word for the shortest/principal/driving recentfile too we should do
660 localroot is good. rfile is good. local_path() is bad, local_path($path)
661 is medium, filenameroot() is bad, remotebase is bad, recentfile is
664 Up to now remotebase was the string that described the remote root
665 directory in rsync notation, like pause.perl.org::authors. And
666 recentfile_basename was "RECENT-1h.yaml".
668 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
670 * The test that was added in today's checkin is a good start for a test
671 of rmirror. We should have more methods in Recent.pm: verify,
672 addmissingfiles. We should verify the current tree, then rmirror it and
673 then verifytree the copy. We could then add some arbitrary file and let
674 it be discovered by addmissingfiles, then rmirror again and then
675 verifytree the copy again.
677 Then we could start stealing from csync2 sqlite database [no port to
678 OSX!] and fill a local DB. And methods to compare the database with the
679 recentfiles. Our strength is that in principle we could maintain state
680 with a single float. We have synced up to 1234567890.123456. If the Z
681 file does not add new files all we have to do is mirror the new ones and
684 This makes it clear that we should extend current protocol and declare
685 that we cheat when we add files too late, just to help the other end
686 keeping track. Ah yes, that's what was meant when zloop was mentioned
689 Maybe need to revisit File::Mirror to help me with this task.
691 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
693 * There must be an allow-me-to-truncate flag in every recentfile.
694 Without it one could construct a sequence of updates winning the locking
695 battle against the aggregator. Only if an aggregator has managed to
696 merge data over to the next level, truncating can be allowed. DONE with
699 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
701 * We should probably guarantee that no duplicates enter the aggregator
704 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
706 * To get merge operation faster would need a good benchmark test. What
707 02 spits out isn't reliable enough and is dominated by many other
710 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
711 Author: Andreas J. Koenig <k@k75.(none)>
712 Date: Sat Aug 2 07:58:04 2008 +0200
716 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
717 Author: Andreas J. Koenig <k@k75.(none)>
718 Date: Sat Aug 2 11:40:29 2008 +0200
720 there was a lot of trying but the effect is hardly measurable with
723 * overhead of connecting seems high. When setting
724 max_files_per_connection to 1 we see that.
726 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
728 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
730 25138.2215600014/3600 = 6.98283932222261
732 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
733 be a bug in the aggregator. FIXED
735 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
737 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
738 we have it, can we mirror it, mirror it!
740 I fear this needs a new class which might be called
741 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
742 and treat them as an entity. I realize that a single recentfile may be
743 sufficient for certain tasks and that it is handy for the low level
744 programmer but it is not nice to use. If there is a delete in the 1h
745 file then the 6h file still contains it. Seekers of the best information
746 need to combine at least some of the recentfiles most of the time.
748 There is the place for the Z loop!
750 But the combination is something to collect in a database, isn't it. Did
751 csync2 just harrumph?
753 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
755 * it just occurred to me that hosts in the same mirroring pool could
756 help out each other even without rewriting the recentfile. Just fetch
757 the stuff to mirror from several places, bingo. But that's something
758 that should rather live in a separate package or in rsync directly.
760 * cronjobs are unsuited because with ntp they would all come at the full
761 minute and disturb each other. Besides that I'd hate to have a backbone
762 with more than a few seconds latency.
764 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
766 * a second rsync server with access control for PAUSE. Port? 873 is the
767 standard port, let's take 8873.
769 * if there were a filesystem based on this, it would have a slow access
770 to inexistent files. It would probably provide wrong readdir (only based
771 on current content) or also a slow one (based on a recentfile written
772 after the call). But it would provide fast access to existing files. Or
773 one would deliberately allow slightly blurred answers based on some
774 sqlite reflection of the recentfiles.
776 * todo: write a variant of mirror() that combines two or more
777 recentfiles and treats them like one
779 * todo: signal handler to remove the tempfile
781 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
783 * now that we have the symlink I forgot how it should be used in
786 * the z loop: add missing files to Z file. Just append them (instead of
787 prepending). So one guy prepends something from the Y file from time to
788 time and another guy appends something rather frequently. Collecting
789 pond. When Y merges into Z, things get epoch and the collecting pond
790 gets smaller. What exactly are "missing files"?
792 take note of current epoch of the alpha file, let's call it the
795 find all files on disk
797 remove all files registered in the recentworld up to recent-ts
799 remove all files that have been deleted after recent-ts according to
802 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
804 * rersyncrecent might be a cronjob with a (locked) state file which
805 contains things like after and maybe last z sync or such?
807 rrr-mirror might be an alternative name but how would we justify the
808 three Rs when there is no Re-Rsync-Recent?
810 With the --loop parameter it is an endless loop, without it is no loop.
811 At least this is simple.
813 * todo: new accssor z-interval specifies how often the Z file is updated
814 against the filesystem. We probably want no epoch stamp on these
815 entries. And we want to be able to filter the entries (e.g. no
816 by-modules and by-category tree)
818 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
820 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
821 into Z. We do not need the other files filled up with filesystem
824 * need interface to query for a file in order to NOT call update on
825 PAUSE a second time within a short time.
827 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
829 * recommended update interval? Makes no sense, is different for
836 change-log-default-name: "Todo"