| blob | 393a205e90d8bbfc914b1a2323e164c2c2a1b118 |
1 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
4 left locks around. DONE
6 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
8 * maybe extend the _overview so that it always says if and where the
9 last file is in the next file and where the next event in the next rf
10 would lie. No, don't like this anymore. REJECT
12 * take the two new redundant tests out again, only the third must
13 survive. DONE
15 * Todo: add a sanity check if the merged structure is really pointing to
16 a different rf and that this different rf is larger. DONE
18 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
20 * now test, if they are overlapping. And test if there is a file in the
21 next rf that would fit into this rf's interval.
23 1h 1222324012.8474 1222322541.7963 0.4086
24 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
25 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
26 1W 1222313218.3626 1221708477.5829 167.9835
28 I suspect that somebody writes a merged timestamp without having merged
29 and then somebody else relies on it.
31 If aggregate is running, the intervals must not be extravagated, if it
32 is not running, there must not be bounds, the total number of events in
33 the system must be counted and must be controlled throughout the tests.
34 That the test required the additional update was probably nonsense,
35 because aggregate can cut pieces too. FIXED & DONE
37 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
39 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
40 changed. FIXED
42 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
44 * Most apparent bug at the moment is that the recentfiles are fetched
45 too often. Only the principal should be fetched and if it has not
46 changed, the others should not be refetched. ATM I must admit that I'm
47 happy that we refetch more often than needed because I can more easily
48 fix bugs while the thing is running.
50 * Let's say, 1220474966.19501 is a timestamp of a file that is already
51 done but the done system does not know about it. The reason for the
52 failure is not known and we never reach the status uptodate because of
53 this. We must get over it.
55 Later it turns out that the origin server had a bug somewhere.
56 1220474966.19042 came after 1220474966.19501. Or better: it was in the
57 array of the recentfile one position above. The bug was my own.
59 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
61 * thinking once again about deletes. The one case is that we still have
62 the "new" and the "delete" in different files. If we pass by the DELETE
63 first which is the usual case we could, for example, keep the path in
64 memory and when we later reach the NEW, we could mark it as done and
65 immediately forget the DELETE. But it would be polite if upstream would
66 tell us the epoch of the NEW within the DELETE record, then we could
67 simply use the DONE mechanism without tracking path names. Ah, but the
68 slave can look it up all on his own account and instead of remembering
69 a name, simply put the timestamp of the NEW file into the DONE system.
71 That's all fine. But there is still the race condition where the server
72 does a delete and the slave does not yet know. So for this time window
73 we must be more tolerant against failure. If we cannot download a file,
74 we should just skip it and should not retry immediately. The whole
75 system should discover the lost thing later. Keeping track with the DONE
76 system should really be a no brainer.
78 But there is something more: the whole filesystem is a database and the
79 recentfiles are one possible representation of it. It's a pretty useful
80 representation I think that's why I have implemented something around
81 it. But for strictly local operation it has little value. For local
82 operation we would much rather have a database. So we would enter every
83 recentfile reading and every rsync operation and for every file the last
84 state change and what it leads to. Then we would always ignore older
85 records without the efforts involved with recentfiles.
87 The database would have: path,recentepoch,rsyncedon,deletedon
89 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
91 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
92 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
94 * statusfile, maybe only on demand, alone to have a sharp debugging
95 tool. It is locked and all recentfiles dump themselves into it and we
96 can build a viewer that lets us know where we stand and what's inside.
98 * how, exactly, do we have to deal with deletes? With rsync errors?
100 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
101 authors) failed: No such file or directory (2)
103 The file above is a delete in 1h and a new in file 1M and the
104 delete in the locally running rmirror did not get propagated to the 1M
105 object. Bug. And the consequence is a standstill.
107 It seems that a slave that works with a file below the principal needs
108 to merge things all the way up to get rid of later deletes. Or keep
109 track of all deletes and skip them later. So we need a trackdeletes.pm
110 similar to the done.pm?
112 * consider the effect when resyncing the recentfile takes longer than
113 the time per loop. Then we never rsync any file. We need to diagnose
114 that and force an increase of that loop time. But when we later are fast
115 enough again because the net has recovered, then we need to switch back
116 to original parameters.
118 * remember to verify that no temp files are left lying around and the
119 signal handler
121 * remember: only the principal recentfile needs expiration, all others
122 shall be expired by principal if it discovers that something has move
123 upstream.
125 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
127 * Always check if we stringify to a higher value than in the entry
128 before.
130 * And in covered make an additional check if we would be able to see a
131 numerical difference between the two numbers and if we can't then switch
132 to a different, more expensive algorithm. Do not want to be caught by
133 floating surprises.
135 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
137 * chaching has several aspects here: we can cache the interval of the
138 recentfile which only will change when the mtime of the file changes. We
139 must re-mirror the recentfile when its ttl has expired. Does have_read
140 tell you anything? It counts nothing at all. Only the mtime is
141 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
142 thing the have_mirrored because it is unrelated to the mtime.
144 * Robustness of floating point calculations! I always thought that the
145 string calculated by the origin server for the floating representation
146 of the epoch time is just a string. When we convert it to a number and
147 later back to a string, the other computer might come to a different
148 conclusion. This must not happen, we want to preserve it unter any
149 circumstances. I will have to write tests with overlong sequences that
150 get lost in arithmetic and must see if all still works well.
152 But one fragile point remains: if one host considers a>b and the other
153 one considers them == but no eq. To prevent this, we must probably do
154 some extra homework.
156 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
158 * the concept of tracking DONE needs an object per recentfile that has
159 something like these methods:
161 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
163 covered() register() covered()
165 The unclear thing is how we translate points in time into intervals. We
166 could pass a reference to the current recent_events array when running
167 we_have(xxx) and let the DONE object iterate over it such that it only
168 has to store a list of intervals that can melt into each other. Ah, even
169 passing the list together with a list of indexes seems feasiable.
171 Or maybe ask for the inverted list?
173 Whenever the complete array is covered by the interval we say we are
174 fully covered and if the recentfile is not expired, we are uptodate.
176 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
178 * idea: have a new flag on recentfiles with the meaning: if this
179 changes, you're required to run a full rsync over all the files. The
180 reason why we set it would probably be: some foul happened. we injected
181 files in arbitrary places or didn't inject them although they changed.
182 The content of the flag? Timestamp? The relation between the
183 recentfiles would have to be inheritance from the principal, because any
184 out of band changes would soon later propagate to the next recentfile.
186 By upping the flag often one can easily ruin the slaves.
188 last out of band change? dirtymark?
190 Anyway, this implies that we read a potentially existing recentfile
191 before we write one.
193 And it implies that we have an eventloop that keeps us busy in 2-3
194 cycles, one for current stuff (tight loop) and one for the recentfiles
195 (cascade when principal has changed), one for the old stuff after a
196 dirtymark change.
198 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
200 * need a way to "return" the next entry after the end of a list. When
201 the caller says "before" or "after" we would like to know if he could
202 cover that interval/threshold or not because this influences the effect
203 of a newer timestamp of that recentfile. DONE with $opt{info}.
205 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
207 * one of the next things to tackle: the equivalent of csync2 -TIXU.
209 loop implies tixu (?). Nope, something like --statefile decides. Per
210 default we do ...?
212 T test, I init, X including removals, U nodirtymark
214 So we have no concept of dirtymarks, we only trust that since we are
215 running we have observed everything steadily. But people will not let
216 this program run forever so we must consider both startup penalty and
217 book keeping for later runs. We keep this for later. For now we write a
218 long running mirror that merges several intervals.
220 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
222 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
223 the intervals!
225 * rersyncrecent, the script: default to one week. The name of the switch
226 is --after. Other switches? --loop!
228 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
230 * need a switch --skip-deletes (?)
232 * need a switch --enduser that tells us that the whole tempfile
233 discipline is not needed when there is no downstream user. (?)
235 Without this switch we cannot have a reasonable recent.pl that just
236 displays the recent additions. Either we accept to download everything.
237 Or we download temporary files without the typical rsync protocol
238 advantages.
240 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
241 File::Temp::tempdir, this might be a win.
243 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
245 * apropos missing: we have no push, we never know the downstream
246 servers. People who know their downstream hosts and want to ascertain
247 something will want additional methods we have never thought about, like
248 update or delete a certain file.
250 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
252 * tempted to refactor rmirror into resolve_symlink, localize, etc.
253 Curious if rsync_options=links equal 0 vs. 1 will make the expected
254 difference.
256 * rsync options: it's a bit of a pain that we usually need several rsync
257 options, like compress, links, times, checksum and that there is no
258 reasonable default except the original rsync default. I think wee can
259 safely assume that the rsync options are shared between all recentfile
260 instances within one recent tree.
262 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
264 * deletes: if a delete follows an add quickly enough it may happen that
265 a downstream mirror did not see the add at all! It seems this needs to
266 be mentioned somewhere. The point here is that even if the downstream is
267 never missing the principal timeframe it may encounter a "delete" that
268 has no complimentary "add" anywhere.
270 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
272 * I suspect the treat of metadata is incorrect during read or something.
273 The bug that I am watching is that between 06:08 and 06:09 the 6h file
274 contained more than 6 hours worth of data. At 06:08 we merged into the
275 1d file. We need to take snapshots of the 6h file over the course of an
276 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
277 enough.
279 Much worse: watching the 1h file: right at the moment (at 06:35) it
280 covers 1218867584-1219120397 which is 70 hours.
282 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
283 08:19:44 2008, that is when I checked out last time, so it seems to be
284 aggregating and never truncating?
286 No, correct is: it is never truncating; but wrong is: it is aggregating.
287 It does receive a lot of events from time to time from a larger file.
288 Somehow a large file gets merged into the small one and because the
289 "meta/merged" attribute is missing, nobody is paying attention. I
290 believe that I can fix this by making sure that metadata are honoured
291 during read. DONE and test adjusted.
293 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
295 * grand renaming plan
297 remotebase => remoteroot to fit well with localroot DONE
298 local_path() => localroot seems to me should already work DONE
299 recentfile_basename => rfilename no need to stress it has no slash DONE
301 filenameroot??? Doesn't seem too bad to me today. Maybe something like
302 kern? It would anyway need a deprecation cycle because it is an
303 important constructor.
305 * I like the portability that Data::Serializer brings us but the price
306 is that some day we might find out that it is slowing us a bit. We'll
307 see.
309 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
311 * should we not enter the interval of the principal (or the interval of
312 the merging file?) in every aggregated/merged file?
314 * we should aim at a first release and give up on thinking about
315 sanitizing stuff and zloop. Let's just admit that a full traditional
316 rsync is the only available sanitizer ATM. Otherwise it's complicated
317 stuff: sanitizing on the origin server, sanitizing on the slaves,
318 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
319 get the basics out before this becomes a major cause for mess.
321 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
323 * On OSes not supporting symlinks we expect that RECENT.recent contains
324 the contents of the principal recentfile. Actually this is identical on
325 systems supporting symlinks. Simple, what follows from that is that we
326 need to keep the serializer in the metadata because we cannot read it
327 from the filename, doesn't it? Of course not. It's a chicken and egg
328 problem. This leaves us with the problem to actually parse the
329 serialized data to find out in which format it is. So who can do the 4
330 or 5 magics we wanted to support? File::LibMagic?
332 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
334 * remotebase and recentfile_basename are ugly names. Now that we need a
335 word for the shortest/principal/driving recentfile too we should do
336 something about it.
338 localroot is good. rfile is good. local_path() is bad, local_path($path)
339 is medium, filenameroot() is bad, remotebase is bad, recentfile is
340 already deprecated.
342 Up to now remotebase was the string that described the remote root
343 directory in rsync notation, like pause.perl.org::authors. And
344 recentfile_basename was "RECENT-1h.yaml".
346 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
348 * The test that was added in today's checkin is a good start for a test
349 of rmirror. We should have more methods in Recent.pm: verify,
350 addmissingfiles. We should verify the current tree, then rmirror it and
351 then verifytree the copy. We could then add some arbitrary file and let
352 it be discovered by addmissingfiles, then rmirror again and then
353 verifytree the copy again.
355 Then we could start stealing from csync2 sqlite database [no port to
356 OSX!] and fill a local DB. And methods to compare the database with the
357 recentfiles. Our strength is that in principle we could maintain state
358 with a single float. We have synced up to 1234567890.123456. If the Z
359 file does not add new files all we have to do is mirror the new ones and
360 delete the goners.
362 This makes it clear that we should extend current protocol and declare
363 that we cheat when we add files too late, just to help the other end
364 keeping track. Ah yes, that's what was meant when zloop was mentioned
365 earlier.
367 Maybe need to revisit File::Mirror to help me with this task.
369 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
371 * There must be an allow-me-to-truncate flag in every recentfile.
372 Without it one could construct a sequence of updates winning the locking
373 battle against the aggregator. Only if an aggregator has managed to
374 merge data over to the next level, truncating can be allowed. DONE with
375 accessor merged.
377 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
379 * We should probably guarantee that no duplicates enter the aggregator
380 array.
382 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
384 * To get merge operation faster would need a good benchmark test. What
385 02 spits out isn't reliable enough and is dominated by many other
386 things. Between
388 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
389 Author: Andreas J. Koenig <k@k75.(none)>
390 Date: Sat Aug 2 07:58:04 2008 +0200
392 and
394 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
395 Author: Andreas J. Koenig <k@k75.(none)>
396 Date: Sat Aug 2 11:40:29 2008 +0200
398 there was a lot of trying but the effect is hardly measurable with
399 current tests.
401 * overhead of connecting seems high. When setting
402 max_files_per_connection to 1 we see that.
404 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
406 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
408 25138.2215600014/3600 = 6.98283932222261
410 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
411 be a bug in the aggregator. FIXED
413 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
415 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
416 we have it, can we mirror it, mirror it!
418 I fear this needs a new class which might be called
419 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
420 and treat them as an entity. I realize that a single recentfile may be
421 sufficient for certain tasks and that it is handy for the low level
422 programmer but it is not nice to use. If there is a delete in the 1h
423 file then the 6h file still contains it. Seekers of the best information
424 need to combine at least some of the recentfiles most of the time.
426 There is the place for the Z loop!
428 But the combination is something to collect in a database, isn't it. Did
429 csync2 just harrumph?
431 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
433 * it just occurred to me that hosts in the same mirroring pool could
434 help out each other even without rewriting the recentfile. Just fetch
435 the stuff to mirror from several places, bingo. But that's something
436 that should rather live in a separate package or in rsync directly.
438 * cronjobs are unsuited because with ntp they would all come at the full
439 minute and disturb each other. Besides that I'd hate to have a backbone
440 with more than a few seconds latency.
442 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
444 * a second rsync server with access control for PAUSE. Port? 873 is the
445 standard port, let's take 8873.
447 * if there were a filesystem based on this, it would have a slow access
448 to inexistent files. It would probably provide wrong readdir (only based
449 on current content) or also a slow one (based on a recentfile written
450 after the call). But it would provide fast access to existing files. Or
451 one would deliberately allow slightly blurred answers based on some
452 sqlite reflection of the recentfiles.
454 * todo: write a variant of mirror() that combines two or more
455 recentfiles and treats them like one
457 * todo: signal handler to remove the tempfile
459 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
461 * now that we have the symlink I forgot how it should be used in
462 practice.
464 * the z loop: add missing files to Z file. Just append them (instead of
465 prepending). So one guy prepends something from the Y file from time to
466 time and another guy appends something rather frequently. Collecting
467 pond. When Y merges into Z, things get epoch and the collecting pond
468 gets smaller. What exactly are "missing files"?
470 take note of current epoch of the alpha file, let's call it the
471 recent-ts
473 find all files on disk
475 remove all files registered in the recentworld up to recent-ts
477 remove all files that have been deleted after recent-ts according to
478 recentworld
480 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
482 * rersyncrecent might be a cronjob with a (locked) state file which
483 contains things like after and maybe last z sync or such?
485 rrr-mirror might be an alternative name but how would we justify the
486 three Rs when there is no Re-Rsync-Recent?
488 With the --loop parameter it is an endless loop, without it is no loop.
489 At least this is simple.
491 * todo: new accssor z-interval specifies how often the Z file is updated
492 against the filesystem. We probably want no epoch stamp on these
493 entries. And we want to be able to filter the entries (e.g. no
494 by-modules and by-category tree)
496 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
498 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
499 into Z. We do not need the other files filled up with filesystem
500 contents though.
502 * need interface to query for a file in order to NOT call update on
503 PAUSE a second time within a short time.
505 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
507 * recommended update interval? Makes no sense, is different for
508 different users.
510 * Moosify
512 Local Variables:
513 mode: change-log
514 change-log-default-name: "Todo"
515 tab-width: 2
516 left-margin: 2
517 End: