| blob | 38fe17e71eb1c8148c586a341c44647806755955 |
1 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * revisit all $rfs->[$i+1] places if they now make sense still
5 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
7 * another bug is the fact that the mirror command deletes files before
8 it unhides the index file, thus confusing downstream slaves. We must not
9 delete before unhiding and must delete after unhiding. FIXED.
11 * new complication about the slave that is playing a server role.
12 Currently we mirror from newest to oldest with a hidden temporary file
13 as index. And when one file is finished, we unhide the index file.
14 Imagine the cascading server/slave is dead for a day. It then starts
15 mirroring again with the freshest thing and unhides the freshest index
16 file when it has worked through it. In that moment it exposes a time
17 hole. Because it now works on the second recentfile which is still
18 hidden.
20 We currently do nothing special to converge after such a drop out. At
21 least not intentionally and robustly and thought through.
23 The algorithm we use to seed the next file needs quite a lot of more
24 robustness than it currently has. Something to do with looking at the
25 merged element of the next rf and when it has dropped off, we seed
26 immediately. And if it ramains dropped off, we seed again, of course.
28 Nope, looking from smaller to larger RFS we look at the merged element
29 of this RF and at the minmax/max element of the next RF. If that
30 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
31 seeding it.
33 And we need a public accessor seed and unseed or seeded. But not the mix
34 of public and private stuff that then is used behind the back.
36 And then the secondary* stuff must go.
38 And we must understand what the impact is on the DONE system. Can it go
39 unnoticed that there was a hole? And could the DONE system have decided
40 the hole is covered? This should be testable with three directories where
41 the middle stops working for a while. Done->merge is suspicious, we must
42 stop it from merging non-conflatable neighbors due to broken continuity.
44 * dirtymark now lives in Recentfile, needs to be used in rmirror.
46 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
48 * Slaven suggests to have the current epoch or the whole current
49 recentfile available from the HTTP server and take it away with
50 keepalive. This direction goes the granularity down to subseconds.
52 We might want to rewrite everything to factor out transport and allow
53 the whole thing to run via HTTP.
55 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
57 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
58 symlink?
60 * smoker on k81 fetching from k75 to verify cascading works. See
61 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
63 * maybe the loop should wait for CHECKSUMS file after every upload. And
64 CPAN.pm needs to deal with timestamps in the future.
66 * do not forget the dirtymark!
68 Text: have a new flag on recentfiles with the meaning: if this
69 changes, you're required to run a full rsync over all the files. The
70 reason why we set it would probably be: some foul happened. we injected
71 files in arbitrary places or didn't inject them although they changed.
72 The content of the flag? Timestamp? The relation between the
73 recentfiles would have to be inheritance from the principal, because any
74 out of band changes would soon later propagate to the next recentfile.
76 By upping the flag often one can easily ruin the slaves.
78 last out of band change? dirtymark?
80 Anyway, this implies that we read a potentially existing recentfile
81 before we write one.
83 And it implies that we have an eventloop that keeps us busy in 2-3
84 cycles, one for current stuff (tight loop) and one for the recentfiles
85 (cascade when principal has changed), one for the old stuff after a
86 dirtymark change.
88 And it implies that the out-of-band change in any of the recentfiles
89 must have a lock on the principal file and there is the place to set the
90 dirtymark.
92 * after the dirtymark is done: fill up recentfiles with fake (historic)
93 entries; fill up with individual corrections; algorithm maybe to be done
94 with bigfloat so that we can always place something in the middle
95 between two entries. Before we must switch to bigfloat we could try to
96 use Data::Float::nextup to get the.
98 * lookup by epoch and by path and use this ability on the pause to never
99 again register a file twice that doesn't need it.
101 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
102 directly.
104 * start a FAQ, especially quick start guide questions. Also to aid those
105 problematic areas where we have no good solution, like the "links"
106 option to rsync.
108 * wish feedback when we are slow.
110 * reduce mccabe
112 * Remove a few DEBUG statements.
114 * The multiple-rrr way of doing things needs a new option to rmirror,
115 like piecemeal or so. Not urgent because after the first pass through,
116 things run smoothely. It's only ugly during the first pass.
118 * I have the suspicion that the code is broken that decides if the
119 neighboring RF needs to be seeded. I fear when too much time has gone
120 between two calls (in our case more than one hour), it would not seed
121 the neighbor. Of course this will never be noticed, so we need a good
122 test for it.
124 * local/localroot confusion: I currently pass both options but one must
125 do.
127 * accounts for early birds on PAUSE rsync daemon.
129 * hardcoded 20 seconds
131 * who mirrors the index? DOING now.
133 * which CPAN mirrors offer rsync?
135 * visit all XXX, visit all _float places
137 * rename the pathdb stuff, it's too confusing. No idea how.
139 * rrr-inotify, backpan, rrr-register
141 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
143 * current bugs: the pathdb seems to get no reset, the seeding of the
144 secondaryttl stuff seems not to have an effect. Have helped myself with
145 a rand(10), need to fix this back. So not checked in. Does the rand
146 thing even help?
148 The rand thing helps. The secondaryttl stuff was in the wrong line,
149 fixed now.
151 The pathdb stuff was because I called either _pathdb or __pathdb on the
152 wrong object. FIXED now.
154 * It's not so beautiful if we never fetch the recentfiles that are not
155 the principal, even if this is correct behaviour. We really do not need
156 them after we have fetched the whole content.
158 OK, we want a switch for that: secondaryttl DONE
160 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
162 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
163 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
164 it lets 40 follow. FIXED
166 * See that the long running process really only updates the principal
167 file unless it has missed a timespan during which something happened. If
168 nothing happened, it must notice even when it misses the timespan. DONE
170 * we must throw away the pathdb when we have reached the end of Z. From
171 that moment we can have a very small pathdb because the only reason for
172 a pathdb is that we know to ignore old records in old files. We won't
173 need this pathdb again before the next full pass over the data is
174 necessary and then we will rebuild it as we go along. DONE
176 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
178 * I think, Done::register_one is doing wrong in that it does not
179 conflate neighboring pieces. The covered() method cannot do this because
180 it has no recent_events array at hand. But register_one has it and could
181 do it and for some reason misses to do it (sometimes).
183 This means that the three tests I just wrote can probably not survive
184 because they test with an already broken Done structure.
186 The art now is to detect how it happens, then to reproduce, then write a
187 test, then fix it.
189 So from the logfile this is what happens: we have a good interval with
190 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
191 top of it. Locally we now mirror F2 and open a new done interval for it.
192 Then we mirror F1 but this time with the timestamp T1b. And when we then
193 try to close the gap, we do not find T1 but instead something older. We
194 should gladly accept this older piece and this would fix this bug.
196 FIXED
198 * bug to fix: when the 1h file changes while rmirror is running, we do
199 correctly sync the new files but never switch to the 6h file but rather
200 stay in a rather quick loop that fetches the 1h file again and again.
202 Is it possible that we initialize a new object? Or does
203 get_remote_recentfile_as_tempfile overwrite something in myself?
205 Want a new option: _runstatusfile => $file which frequently dumps the
206 state of all recentfiles to a file.
208 FIXED
210 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
212 * Todo: now teach update to verify the timestamp is about to write
213 against the previous and use _increase_a_bit if it doesn't comply with
214 strict monotony. DONE
216 * The problem of rounding. So far perl's default precision was
217 sufficient. One day it won't be. FakeFloat has an easy job when it is
218 only reading and other machines have written correctly. But when we want
219 to write a floating point number that is a bit larger than the other
220 one, then we need our own idea of precision.
222 Slaven said: just append a "1". This might be going towards the end of
223 usability too quickly. I'd like something that actually uses the decimal
224 system. Well, appending a 1 also does this but...
226 E.g. we have 1.0. nextup on this architecture is starting with
227 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
228 taken the 1.0000000000000003 and the next user comes and the time tells
229 him 1.0 again. He has to beat my number without stepping over the
230 nextup. This is much less space than I had when I chose 1,2,3.
232 What is also irritating is that nextup is architecture dependent. The
233 128 bit guy must choose very long numbers to fit in between whereas the
234 other one with 16 bit uses larger steps. But then the algorithm is the
235 same for both, so that would be a nice thing.
237 I see two situation where we need this. One is when Time::HiRes returns
238 us a value that is <= the last entry in our recentfile. In this case
239 (let's call it the end-case) we must fill the region between that number
240 and the next higher native floating point number. The other is when we
241 inject an old file into an old recentfile (we would then also set a new
242 dirtymark). We find the integer value already taken and need a slightly
243 different one (let's call it the middle-case). The difference between
244 the two situations is that the next user will want to find something
245 higher than my number in the end-case and something lower than my number
246 in the middle case.
248 So I suggest we give the function both a value and an upper bound and it
249 calculates us a primitive middle. The upper bound in the middle-case is
250 the next integer. The upper bound on the end-case is the nextup floating
251 point number. But the latter poses another problem: if we have occupied
252 the middle m between x and nextup(x), then the nextup(m) will probably
253 not be the same as nextup(x) because some rounding will take place
254 before the nextup is calculated and when the rounding reaches the
255 nextup(x), we will end up at nextup(nextup(x)).
257 So we really need to consider the nextup and the nextdown from there and
258 then the middle and that's the number we may approach asymptotically.
259 Ugly. But DONE.
261 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
263 * consider deprecating the use of RECENT.recent as a symlink. It turns
264 out to need extra hoops with the rsync options and just isn't worth it.
265 Or maybe these extra hoops are needed anyway for the rest of the tree?
266 Nope, can't be the case because not all filesystems support symlinks.
268 But before doing the large step, I'll deprecate the call of
269 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
270 only introduced to resolve RECENT.recent and complicates the routine far
271 beyond what it deserves.
273 DONE. Won't deprecate RECENT.recent, just moved its handling to the
274 supervisor.
276 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
278 * I think it's a bug that the rsync_option links must be set to true in
279 order to support RECENT.recent and that nobody cares to set it
280 automatically. Similar for ignore_link_stat_errors.
282 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
284 * Todo: find all todos together and make a plan what is missing for a
285 release.
287 - verifytree or something like that. fsck maybe.
289 - rersyncrecent, the script itself? What it do?
291 - a way to only mirror the recentfiles without mirroring the whole
292 remote system such that people can decide to mirror only partially see
293 also 2008-08-30. .shadow-xxx directory? this also needed for a
294 filesystem that is still incomplete and might need the mirrorfiles for
295 lookup(?)
297 - long living objects that mirror again and again. Inject something
298 into ta, see how it goes over to tb.
300 - how do we continue filling up the DONE system when we use an object
301 for the second time? "fully covered" and "uptodate" or new terminology.
303 - overview called on the wrong file should be understandable
305 - the meta data field that must change when we fake something up so that
306 the downstream people know they have to re-fetch everything.
308 - how tolerant are we against missing files upstream? how do we keep
309 track? there are legitimate cases where we did read upstream index right
310 before a file got deleted there and then find that file as new and want
311 it. There are other cases that are not self healing and must be tracked
312 and bugreported.
314 - how, exactly, do we have to deal with deletes? With rsync errors?
316 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
317 authors) failed: No such file or directory (2)
319 The file above is a delete in 1h and a new in file 1M and the
320 delete in the locally running rmirror did not get propagated to the 1M
321 object. Bug. And the consequence is a standstill.
323 It seems that a slave that works with a file below the principal needs
324 to merge things all the way up to get rid of later deletes. Or keep
325 track of all deletes and skip them later. So we need a trackdeletes.pm
326 similar to the done.pm?
328 see also 2008-08-20 about spurious deletes that really have no add
329 counterpart and yet they are not wrong.
331 - consider the effect when resyncing the recentfile takes longer than
332 the time per loop. Then we never rsync any file. We need to diagnose
333 that and force an increase of that loop time. But when we later are fast
334 enough again because the net has recovered, then we need to switch back
335 to original parameters. ERm, no, it's enough to keep syncing at least
336 one file before refetching an index file.
338 - remember to verify that no temp files are left lying around and the
339 signal handler
341 - status file for not long running jobs that want to track upstream with
342 a, say, cronjob.
344 - revisit all XXX _float areas and study Sub::Exporter DONE
346 - persistent DB even though we just said we do not need it. Just for
347 extended capabilities and time savings when, for example, upstream
348 announces a reset and we get new recentfiles and could then limit
349 ourselves to a subset of files (those that have a changed epoch) in a
350 first pass and would only then do the loop to verify the rest. Or
351 something.
353 * Todo: aggregate files should know their feed and finding the principal
354 should be done stepwise. (?)
356 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
357 left locks around. DONE
359 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
361 * maybe extend the _overview so that it always says if and where the
362 last file is in the next file and where the next event in the next rf
363 would lie. No, don't like this anymore. REJECT
365 * take the two new redundant tests out again, only the third must
366 survive. DONE
368 * Todo: add a sanity check if the merged structure is really pointing to
369 a different rf and that this different rf is larger. DONE
371 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
373 * now test, if they are overlapping. And test if there is a file in the
374 next rf that would fit into this rf's interval.
376 1h 1222324012.8474 1222322541.7963 0.4086
377 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
378 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
379 1W 1222313218.3626 1221708477.5829 167.9835
381 I suspect that somebody writes a merged timestamp without having merged
382 and then somebody else relies on it.
384 If aggregate is running, the intervals must not be extravagated, if it
385 is not running, there must not be bounds, the total number of events in
386 the system must be counted and must be controlled throughout the tests.
387 That the test required the additional update was probably nonsense,
388 because aggregate can cut pieces too. FIXED & DONE
390 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
392 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
393 changed. FIXED
395 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
397 * Most apparent bug at the moment is that the recentfiles are fetched
398 too often. Only the principal should be fetched and if it has not
399 changed, the others should not be refetched. ATM I must admit that I'm
400 happy that we refetch more often than needed because I can more easily
401 fix bugs while the thing is running.
403 * Let's say, 1220474966.19501 is a timestamp of a file that is already
404 done but the done system does not know about it. The reason for the
405 failure is not known and we never reach the status uptodate because of
406 this. We must get over it.
408 Later it turns out that the origin server had a bug somewhere.
409 1220474966.19042 came after 1220474966.19501. Or better: it was in the
410 array of the recentfile one position above. The bug was my own.
412 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
414 * There is the race condition where the server does a delete and the
415 slave does not yet know and then tries to download it because he sees
416 the new. So for this time window we must be more tolerant against
417 failure. If we cannot download a file, we should just skip it and should
418 not retry immediately. The whole system should discover the lost thing
419 later. Keeping track with the DONE system should really be a no brainer.
421 But there is something more: the whole filesystem is a database and the
422 recentfiles are one possible representation of it. It's a pretty useful
423 representation I think that's why I have implemented something around
424 it. But for strictly local operation it has little value. For local
425 operation we would much rather have a database. So we would enter every
426 recentfile reading and every rsync operation and for every file the last
427 state change and what it leads to. Then we would always ignore older
428 records without the efforts involved with recentfiles.
430 The database would have: path,recentepoch,rsyncedon,deletedon
432 Oh well, not yet clear where this leads to.
434 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
436 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
437 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
439 * statusfile, maybe only on demand, alone to have a sharp debugging
440 tool. It is locked and all recentfiles dump themselves into it and we
441 can build a viewer that lets us know where we stand and what's inside.
443 * remember: only the principal recentfile needs expiration, all others
444 shall be expired by principal if it discovers that something has move
445 upstream.
447 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
449 * Always check if we stringify to a higher value than in the entry
450 before. DONE
452 * And in covered make an additional check if we would be able to see a
453 numerical difference between the two numbers and if we can't then switch
454 to a different, more expensive algorithm. Do not want to be caught by
455 floating surprises. DONE
457 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
459 * caching has several aspects here: we can cache the interval of the
460 recentfile which only will change when the mtime of the file changes. We
461 must re-mirror the recentfile when its ttl has expired. Does have_read
462 tell you anything? It counts nothing at all. Only the mtime is
463 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
464 thing the have_mirrored because it is unrelated to the mtime.
466 * Robustness of floating point calculations! I always thought that the
467 string calculated by the origin server for the floating representation
468 of the epoch time is just a string. When we convert it to a number and
469 later back to a string, the other computer might come to a different
470 conclusion. This must not happen, we want to preserve it under any
471 circumstances. I will have to write tests with overlong sequences that
472 get lost in arithmetic and must see if all still works well. DONE
474 But one fragile point remains: if one host considers a>b and the other
475 one considers them == but no eq. To prevent this, we must probably do
476 some extra homework. DONE
478 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
480 * the concept of tracking DONE needs an object per recentfile that has
481 something like these methods:
483 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
485 covered() register() covered()
487 The unclear thing is how we translate points in time into intervals. We
488 could pass a reference to the current recent_events array when running
489 we_have(xxx) and let the DONE object iterate over it such that it only
490 has to store a list of intervals that can melt into each other. Ah, even
491 passing the list together with a list of indexes seems feasiable.
493 Or maybe ask for the inverted list?
495 Whenever the complete array is covered by the interval we say we are
496 fully covered and if the recentfile is not expired, we are uptodate.
498 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
500 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
502 * need a way to "return" the next entry after the end of a list. When
503 the caller says "before" or "after" we would like to know if he could
504 cover that interval/threshold or not because this influences the effect
505 of a newer timestamp of that recentfile. DONE with $opt{info}.
507 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
509 * one of the next things to tackle: the equivalent of csync2 -TIXU.
511 loop implies tixu (?). Nope, something like --statefile decides. Per
512 default we do ...?
514 T test, I init, X including removals, U nodirtymark
516 So we have no concept of dirtymarks, we only trust that since we are
517 running we have observed everything steadily. But people will not let
518 this program run forever so we must consider both startup penalty and
519 book keeping for later runs. We keep this for later. For now we write a
520 long running mirror that merges several intervals.
522 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
524 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
525 the intervals!
527 * rersyncrecent, the script: default to one week. The name of the switch
528 is --after. Other switches? --loop!
530 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
532 * need a switch --skip-deletes (?)
534 * need a switch --enduser that tells us that the whole tempfile
535 discipline is not needed when there is no downstream user. (?)
537 Without this switch we cannot have a reasonable recent.pl that just
538 displays the recent additions. Either we accept to download everything.
539 Or we download temporary files without the typical rsync protocol
540 advantages.
542 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
543 File::Temp::tempdir, this might be a win.
545 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
547 * apropos missing: we have no push, we never know the downstream
548 servers. People who know their downstream hosts and want to ascertain
549 something will want additional methods we have never thought about, like
550 update or delete a certain file.
552 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
554 * tempted to refactor rmirror into resolve_symlink, localize, etc.
555 Curious if rsync_options=links equal 0 vs. 1 will make the expected
556 difference.
558 * rsync options: it's a bit of a pain that we usually need several rsync
559 options, like compress, links, times, checksum and that there is no
560 reasonable default except the original rsync default. I think wee can
561 safely assume that the rsync options are shared between all recentfile
562 instances within one recent tree.
564 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
566 * deletes: if a delete follows an add quickly enough it may happen that
567 a downstream mirror did not see the add at all! It seems this needs to
568 be mentioned somewhere. The point here is that even if the downstream is
569 never missing the principal timeframe it may encounter a "delete" that
570 has no complimentary "add" anywhere.
572 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
574 * I suspect the treat of metadata is incorrect during read or something.
575 The bug that I am watching is that between 06:08 and 06:09 the 6h file
576 contained more than 6 hours worth of data. At 06:08 we merged into the
577 1d file. We need to take snapshots of the 6h file over the course of an
578 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
579 enough.
581 Much worse: watching the 1h file: right at the moment (at 06:35) it
582 covers 1218867584-1219120397 which is 70 hours.
584 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
585 08:19:44 2008, that is when I checked out last time, so it seems to be
586 aggregating and never truncating?
588 No, correct is: it is never truncating; but wrong is: it is aggregating.
589 It does receive a lot of events from time to time from a larger file.
590 Somehow a large file gets merged into the small one and because the
591 "meta/merged" attribute is missing, nobody is paying attention. I
592 believe that I can fix this by making sure that metadata are honoured
593 during read. DONE and test adjusted.
595 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
597 * grand renaming plan
599 remotebase => remoteroot to fit well with localroot DONE
600 local_path() => localroot seems to me should already work DONE
601 recentfile_basename => rfilename no need to stress it has no slash DONE
603 filenameroot??? Doesn't seem too bad to me today. Maybe something like
604 kern? It would anyway need a deprecation cycle because it is an
605 important constructor.
607 * I like the portability that Data::Serializer brings us but the price
608 is that some day we might find out that it is slowing us a bit. We'll
609 see.
611 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
613 * should we not enter the interval of the principal (or the interval of
614 the merging file?) in every aggregated/merged file?
616 * we should aim at a first release and give up on thinking about
617 sanitizing stuff and zloop. Let's just admit that a full traditional
618 rsync is the only available sanitizer ATM. Otherwise it's complicated
619 stuff: sanitizing on the origin server, sanitizing on the slaves,
620 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
621 get the basics out before this becomes a major cause for mess.
623 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
625 * On OSes not supporting symlinks we expect that RECENT.recent contains
626 the contents of the principal recentfile. Actually this is identical on
627 systems supporting symlinks. Simple, what follows from that is that we
628 need to keep the serializer in the metadata because we cannot read it
629 from the filename, doesn't it? Of course not. It's a chicken and egg
630 problem. This leaves us with the problem to actually parse the
631 serialized data to find out in which format it is. So who can do the 4
632 or 5 magics we wanted to support? File::LibMagic?
634 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
636 * remotebase and recentfile_basename are ugly names. Now that we need a
637 word for the shortest/principal/driving recentfile too we should do
638 something about it.
640 localroot is good. rfile is good. local_path() is bad, local_path($path)
641 is medium, filenameroot() is bad, remotebase is bad, recentfile is
642 already deprecated.
644 Up to now remotebase was the string that described the remote root
645 directory in rsync notation, like pause.perl.org::authors. And
646 recentfile_basename was "RECENT-1h.yaml".
648 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
650 * The test that was added in today's checkin is a good start for a test
651 of rmirror. We should have more methods in Recent.pm: verify,
652 addmissingfiles. We should verify the current tree, then rmirror it and
653 then verifytree the copy. We could then add some arbitrary file and let
654 it be discovered by addmissingfiles, then rmirror again and then
655 verifytree the copy again.
657 Then we could start stealing from csync2 sqlite database [no port to
658 OSX!] and fill a local DB. And methods to compare the database with the
659 recentfiles. Our strength is that in principle we could maintain state
660 with a single float. We have synced up to 1234567890.123456. If the Z
661 file does not add new files all we have to do is mirror the new ones and
662 delete the goners.
664 This makes it clear that we should extend current protocol and declare
665 that we cheat when we add files too late, just to help the other end
666 keeping track. Ah yes, that's what was meant when zloop was mentioned
667 earlier.
669 Maybe need to revisit File::Mirror to help me with this task.
671 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
673 * There must be an allow-me-to-truncate flag in every recentfile.
674 Without it one could construct a sequence of updates winning the locking
675 battle against the aggregator. Only if an aggregator has managed to
676 merge data over to the next level, truncating can be allowed. DONE with
677 accessor merged.
679 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
681 * We should probably guarantee that no duplicates enter the aggregator
682 array.
684 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
686 * To get merge operation faster would need a good benchmark test. What
687 02 spits out isn't reliable enough and is dominated by many other
688 things. Between
690 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
691 Author: Andreas J. Koenig <k@k75.(none)>
692 Date: Sat Aug 2 07:58:04 2008 +0200
694 and
696 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
697 Author: Andreas J. Koenig <k@k75.(none)>
698 Date: Sat Aug 2 11:40:29 2008 +0200
700 there was a lot of trying but the effect is hardly measurable with
701 current tests.
703 * overhead of connecting seems high. When setting
704 max_files_per_connection to 1 we see that.
706 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
708 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
710 25138.2215600014/3600 = 6.98283932222261
712 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
713 be a bug in the aggregator. FIXED
715 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
717 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
718 we have it, can we mirror it, mirror it!
720 I fear this needs a new class which might be called
721 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
722 and treat them as an entity. I realize that a single recentfile may be
723 sufficient for certain tasks and that it is handy for the low level
724 programmer but it is not nice to use. If there is a delete in the 1h
725 file then the 6h file still contains it. Seekers of the best information
726 need to combine at least some of the recentfiles most of the time.
728 There is the place for the Z loop!
730 But the combination is something to collect in a database, isn't it. Did
731 csync2 just harrumph?
733 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
735 * it just occurred to me that hosts in the same mirroring pool could
736 help out each other even without rewriting the recentfile. Just fetch
737 the stuff to mirror from several places, bingo. But that's something
738 that should rather live in a separate package or in rsync directly.
740 * cronjobs are unsuited because with ntp they would all come at the full
741 minute and disturb each other. Besides that I'd hate to have a backbone
742 with more than a few seconds latency.
744 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
746 * a second rsync server with access control for PAUSE. Port? 873 is the
747 standard port, let's take 8873.
749 * if there were a filesystem based on this, it would have a slow access
750 to inexistent files. It would probably provide wrong readdir (only based
751 on current content) or also a slow one (based on a recentfile written
752 after the call). But it would provide fast access to existing files. Or
753 one would deliberately allow slightly blurred answers based on some
754 sqlite reflection of the recentfiles.
756 * todo: write a variant of mirror() that combines two or more
757 recentfiles and treats them like one
759 * todo: signal handler to remove the tempfile
761 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
763 * now that we have the symlink I forgot how it should be used in
764 practice.
766 * the z loop: add missing files to Z file. Just append them (instead of
767 prepending). So one guy prepends something from the Y file from time to
768 time and another guy appends something rather frequently. Collecting
769 pond. When Y merges into Z, things get epoch and the collecting pond
770 gets smaller. What exactly are "missing files"?
772 take note of current epoch of the alpha file, let's call it the
773 recent-ts
775 find all files on disk
777 remove all files registered in the recentworld up to recent-ts
779 remove all files that have been deleted after recent-ts according to
780 recentworld
782 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
784 * rersyncrecent might be a cronjob with a (locked) state file which
785 contains things like after and maybe last z sync or such?
787 rrr-mirror might be an alternative name but how would we justify the
788 three Rs when there is no Re-Rsync-Recent?
790 With the --loop parameter it is an endless loop, without it is no loop.
791 At least this is simple.
793 * todo: new accssor z-interval specifies how often the Z file is updated
794 against the filesystem. We probably want no epoch stamp on these
795 entries. And we want to be able to filter the entries (e.g. no
796 by-modules and by-category tree)
798 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
800 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
801 into Z. We do not need the other files filled up with filesystem
802 contents though.
804 * need interface to query for a file in order to NOT call update on
805 PAUSE a second time within a short time.
807 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
809 * recommended update interval? Makes no sense, is different for
810 different users.
812 * Moosify
814 Local Variables:
815 mode: change-log
816 change-log-default-name: "Todo"
817 tab-width: 2
818 left-margin: 2
819 End: