| blob | ed582c2096f1e08aaa5cc4504802bf781c07ad7f |
1 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * maybe extend the _overview so that it always says if and where the
4 last file is in the next file and where the next event in the next rf
5 would lie.
7 * take the two new redundant tests out again, only the third must
8 survive. DONE
10 * Todo: add a sanity check if the merged structure is really pointing to
11 a different rf and that this different rf is larger.
13 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
15 * now test, if they are overlapping. And test if there is a file in the
16 next rf that would fit into this rf's interval.
18 1h 1222324012.8474 1222322541.7963 0.4086
19 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
20 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
21 1W 1222313218.3626 1221708477.5829 167.9835
23 I suspect that somebody writes a merged timestamp without having merged
24 and then somebody else relies on it.
26 If aggregate is running, the intervals must not be extravagated, if it
27 is not running, there must not be bounds, the total number of events in
28 the system must be counted and must be controlled throughout the tests.
29 That the test required the additional update was probably nonsense,
30 because aggregate can cut pieces too.
32 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
34 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
35 changed. FIXED
37 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
39 * Most apparent bug at the moment is that the recentfiles are fetched
40 too often. Only the principal should be fetched and if it has not
41 changed, the others should not be refetched. ATM I must admit that I'm
42 happy that we refetch more often than needed because I can more easily
43 fix bugs while the thing is running.
45 * Let's say, 1220474966.19501 is a timestamp of a file that is already
46 done but the done system does not know about it. The reason for the
47 failure is not known and we never reach the status uptodate because of
48 this. We must get over it.
50 Later it turns out that the origin server had a bug somewhere.
51 1220474966.19042 came after 1220474966.19501. Or better: it was in the
52 array of the recentfile one position above. The bug was my own.
54 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
56 * thinking once again about deletes. The one case is that we still have
57 the "new" and the "delete" in different files. If we pass by the DELETE
58 first which is the usual case we could, for example, keep the path in
59 memory and when we later reach the NEW, we could mark it as done and
60 immediately forget the DELETE. But it would be polite if upstream would
61 tell us the epoch of the NEW within the DELETE record, then we could
62 simply use the DONE mechanism without tracking path names. Ah, but the
63 slave can look it up all on his own account and instead of remembering
64 a name, simply put the timestamp of the NEW file into the DONE system.
66 That's all fine. But there is still the race condition where the server
67 does a delete and the slave does not yet know. So for this time window
68 we must be more tolerant against failure. If we cannot download a file,
69 we should just skip it and should not retry immediately. The whole
70 system should discover the lost thing later. Keeping track with the DONE
71 system should really be a no brainer.
73 But there is something more: the whole filesystem is a database and the
74 recentfiles are one possible representation of it. It's a pretty useful
75 representation I think that's why I have implemented something around
76 it. But for strictly local operation it has little value. For local
77 operation we would much rather have a database. So we would enter every
78 recentfile reading and every rsync operation and for every file the last
79 state change and what it leads to. Then we would always ignore older
80 records without the efforts involved with recentfiles.
82 The database would have: path,recentepoch,rsyncedon,deletedon
84 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
86 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
87 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
89 * statusfile, maybe only on demand, alone to have a sharp debugging
90 tool. It is locked and all recentfiles dump themselves into it and we
91 can build a viewer that lets us know where we stand and what's inside.
93 * how, exactly, do we have to deal with deletes? With rsync errors?
95 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
96 authors) failed: No such file or directory (2)
98 The file above is a delete in 1h and a new in file 1M and the
99 delete in the locally running rmirror did not get propagated to the 1M
100 object. Bug. And the consequence is a standstill.
102 It seems that a slave that works with a file below the principal needs
103 to merge things all the way up to get rid of later deletes. Or keep
104 track of all deletes and skip them later. So we need a trackdeletes.pm
105 similar to the done.pm?
107 * consider the effect when resyncing the recentfile takes longer than
108 the time per loop. Then we never rsync any file. We need to diagnose
109 that and force an increase of that loop time. But when we later are fast
110 enough again because the net has recovered, then we need to switch back
111 to original parameters.
113 * remember to verify that no temp files are left lying around and the
114 signal handler
116 * remember: only the principal recentfile needs expiration, all others
117 shall be expired by principal if it discovers that something has move
118 upstream.
120 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
122 * Always check if we stringify to a higher value than in the entry
123 before.
125 * And in covered make an additional check if we would be able to see a
126 numerical difference between the two numbers and if we can't then switch
127 to a different, more expensive algorithm. Do not want to be caught by
128 floating surprises.
130 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
132 * chaching has several aspects here: we can cache the interval of the
133 recentfile which only will change when the mtime of the file changes. We
134 must re-mirror the recentfile when its ttl has expired. Does have_read
135 tell you anything? It counts nothing at all. Only the mtime is
136 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
137 thing the have_mirrored because it is unrelated to the mtime.
139 * Robustness of floating point calculations! I always thought that the
140 string calculated by the origin server for the floating representation
141 of the epoch time is just a string. When we convert it to a number and
142 later back to a string, the other computer might come to a different
143 conclusion. This must not happen, we want to preserve it unter any
144 circumstances. I will have to write tests with overlong sequences that
145 get lost in arithmetic and must see if all still works well.
147 But one fragile point remains: if one host considers a>b and the other
148 one considers them == but no eq. To prevent this, we must probably do
149 some extra homework.
151 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
153 * the concept of tracking DONE needs an object per recentfile that has
154 something like these methods:
156 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
158 covered() register() covered()
160 The unclear thing is how we translate points in time into intervals. We
161 could pass a reference to the current recent_events array when running
162 we_have(xxx) and let the DONE object iterate over it such that it only
163 has to store a list of intervals that can melt into each other. Ah, even
164 passing the list together with a list of indexes seems feasiable.
166 Or maybe ask for the inverted list?
168 Whenever the complete array is covered by the interval we say we are
169 fully covered and if the recentfile is not expired, we are uptodate.
171 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
173 * idea: have a new flag on recentfiles with the meaning: if this
174 changes, you're required to run a full rsync over all the files. The
175 reason why we set it would probably be: some foul happened. we injected
176 files in arbitrary places or didn't inject them although they changed.
177 The content of the flag? Timestamp? The relation between the
178 recentfiles would have to be inheritance from the principal, because any
179 out of band changes would soon later propagate to the next recentfile.
181 By upping the flag often one can easily ruin the slaves.
183 last out of band change? dirtymark?
185 Anyway, this implies that we read a potentially existing recentfile
186 before we write one.
188 And it implies that we have an eventloop that keeps us busy in 2-3
189 cycles, one for current stuff (tight loop) and one for the recentfiles
190 (cascade when principal has changed), one for the old stuff after a
191 dirtymark change.
193 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
195 * need a way to "return" the next entry after the end of a list. When
196 the caller says "before" or "after" we would like to know if he could
197 cover that interval/threshold or not because this influences the effect
198 of a newer timestamp of that recentfile. DONE with $opt{info}.
200 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
202 * one of the next things to tackle: the equivalent of csync2 -TIXU.
204 loop implies tixu (?). Nope, something like --statefile decides. Per
205 default we do ...?
207 T test, I init, X including removals, U nodirtymark
209 So we have no concept of dirtymarks, we only trust that since we are
210 running we have observed everything steadily. But people will not let
211 this program run forever so we must consider both startup penalty and
212 book keeping for later runs. We keep this for later. For now we write a
213 long running mirror that merges several intervals.
215 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
217 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
218 the intervals!
220 * rersyncrecent, the script: default to one week. The name of the switch
221 is --after. Other switches? --loop!
223 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
225 * need a switch --skip-deletes (?)
227 * need a switch --enduser that tells us that the whole tempfile
228 discipline is not needed when there is no downstream user. (?)
230 Without this switch we cannot have a reasonable recent.pl that just
231 displays the recent additions. Either we accept to download everything.
232 Or we download temporary files without the typical rsync protocol
233 advantages.
235 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
236 File::Temp::tempdir, this might be a win.
238 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
240 * apropos missing: we have no push, we never know the downstream
241 servers. People who know their downstream hosts and want to ascertain
242 something will want additional methods we have never thought about, like
243 update or delete a certain file.
245 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
247 * tempted to refactor rmirror into resolve_symlink, localize, etc.
248 Curious if rsync_options=links equal 0 vs. 1 will make the expected
249 difference.
251 * rsync options: it's a bit of a pain that we usually need several rsync
252 options, like compress, links, times, checksum and that there is no
253 reasonable default except the original rsync default. I think wee can
254 safely assume that the rsync options are shared between all recentfile
255 instances within one recent tree.
257 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
259 * deletes: if a delete follows an add quickly enough it may happen that
260 a downstream mirror did not see the add at all! It seems this needs to
261 be mentioned somewhere. The point here is that even if the downstream is
262 never missing the principal timeframe it may encounter a "delete" that
263 has no complimentary "add" anywhere.
265 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
267 * I suspect the treat of metadata is incorrect during read or something.
268 The bug that I am watching is that between 06:08 and 06:09 the 6h file
269 contained more than 6 hours worth of data. At 06:08 we merged into the
270 1d file. We need to take snapshots of the 6h file over the course of an
271 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
272 enough.
274 Much worse: watching the 1h file: right at the moment (at 06:35) it
275 covers 1218867584-1219120397 which is 70 hours.
277 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
278 08:19:44 2008, that is when I checked out last time, so it seems to be
279 aggregating and never truncating?
281 No, correct is: it is never truncating; but wrong is: it is aggregating.
282 It does receive a lot of events from time to time from a larger file.
283 Somehow a large file gets merged into the small one and because the
284 "meta/merged" attribute is missing, nobody is paying attention. I
285 believe that I can fix this by making sure that metadata are honoured
286 during read. DONE and test adjusted.
288 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
290 * grand renaming plan
292 remotebase => remoteroot to fit well with localroot DONE
293 local_path() => localroot seems to me should already work DONE
294 recentfile_basename => rfilename no need to stress it has no slash DONE
296 filenameroot??? Doesn't seem too bad to me today. Maybe something like
297 kern? It would anyway need a deprecation cycle because it is an
298 important constructor.
300 * I like the portability that Data::Serializer brings us but the price
301 is that some day we might find out that it is slowing us a bit. We'll
302 see.
304 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
306 * should we not enter the interval of the principal (or the interval of
307 the merging file?) in every aggregated/merged file?
309 * we should aim at a first release and give up on thinking about
310 sanitizing stuff and zloop. Let's just admit that a full traditional
311 rsync is the only available sanitizer ATM. Otherwise it's complicated
312 stuff: sanitizing on the origin server, sanitizing on the slaves,
313 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
314 get the basics out before this becomes a major cause for mess.
316 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
318 * On OSes not supporting symlinks we expect that RECENT.recent contains
319 the contents of the principal recentfile. Actually this is identical on
320 systems supporting symlinks. Simple, what follows from that is that we
321 need to keep the serializer in the metadata because we cannot read it
322 from the filename, doesn't it? Of course not. It's a chicken and egg
323 problem. This leaves us with the problem to actually parse the
324 serialized data to find out in which format it is. So who can do the 4
325 or 5 magics we wanted to support? File::LibMagic?
327 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
329 * remotebase and recentfile_basename are ugly names. Now that we need a
330 word for the shortest/principal/driving recentfile too we should do
331 something about it.
333 localroot is good. rfile is good. local_path() is bad, local_path($path)
334 is medium, filenameroot() is bad, remotebase is bad, recentfile is
335 already deprecated.
337 Up to now remotebase was the string that described the remote root
338 directory in rsync notation, like pause.perl.org::authors. And
339 recentfile_basename was "RECENT-1h.yaml".
341 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
343 * The test that was added in today's checkin is a good start for a test
344 of rmirror. We should have more methods in Recent.pm: verify,
345 addmissingfiles. We should verify the current tree, then rmirror it and
346 then verifytree the copy. We could then add some arbitrary file and let
347 it be discovered by addmissingfiles, then rmirror again and then
348 verifytree the copy again.
350 Then we could start stealing from csync2 sqlite database [no port to
351 OSX!] and fill a local DB. And methods to compare the database with the
352 recentfiles. Our strength is that in principle we could maintain state
353 with a single float. We have synced up to 1234567890.123456. If the Z
354 file does not add new files all we have to do is mirror the new ones and
355 delete the goners.
357 This makes it clear that we should extend current protocol and declare
358 that we cheat when we add files too late, just to help the other end
359 keeping track. Ah yes, that's what was meant when zloop was mentioned
360 earlier.
362 Maybe need to revisit File::Mirror to help me with this task.
364 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
366 * There must be an allow-me-to-truncate flag in every recentfile.
367 Without it one could construct a sequence of updates winning the locking
368 battle against the aggregator. Only if an aggregator has managed to
369 merge data over to the next level, truncating can be allowed. DONE with
370 accessor merged.
372 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
374 * We should probably guarantee that no duplicates enter the aggregator
375 array.
377 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
379 * To get merge operation faster would need a good benchmark test. What
380 02 spits out isn't reliable enough and is dominated by many other
381 things. Between
383 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
384 Author: Andreas J. Koenig <k@k75.(none)>
385 Date: Sat Aug 2 07:58:04 2008 +0200
387 and
389 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
390 Author: Andreas J. Koenig <k@k75.(none)>
391 Date: Sat Aug 2 11:40:29 2008 +0200
393 there was a lot of trying but the effect is hardly measurable with
394 current tests.
396 * overhead of connecting seems high. When setting
397 max_files_per_connection to 1 we see that.
399 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
401 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
403 25138.2215600014/3600 = 6.98283932222261
405 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
406 be a bug in the aggregator. FIXED
408 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
410 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
411 we have it, can we mirror it, mirror it!
413 I fear this needs a new class which might be called
414 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
415 and treat them as an entity. I realize that a single recentfile may be
416 sufficient for certain tasks and that it is handy for the low level
417 programmer but it is not nice to use. If there is a delete in the 1h
418 file then the 6h file still contains it. Seekers of the best information
419 need to combine at least some of the recentfiles most of the time.
421 There is the place for the Z loop!
423 But the combination is something to collect in a database, isn't it. Did
424 csync2 just harrumph?
426 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
428 * it just occurred to me that hosts in the same mirroring pool could
429 help out each other even without rewriting the recentfile. Just fetch
430 the stuff to mirror from several places, bingo. But that's something
431 that should rather live in a separate package or in rsync directly.
433 * cronjobs are unsuited because with ntp they would all come at the full
434 minute and disturb each other. Besides that I'd hate to have a backbone
435 with more than a few seconds latency.
437 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
439 * a second rsync server with access control for PAUSE. Port? 873 is the
440 standard port, let's take 8873.
442 * if there were a filesystem based on this, it would have a slow access
443 to inexistent files. It would probably provide wrong readdir (only based
444 on current content) or also a slow one (based on a recentfile written
445 after the call). But it would provide fast access to existing files. Or
446 one would deliberately allow slightly blurred answers based on some
447 sqlite reflection of the recentfiles.
449 * todo: write a variant of mirror() that combines two or more
450 recentfiles and treats them like one
452 * todo: signal handler to remove the tempfile
454 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
456 * now that we have the symlink I forgot how it should be used in
457 practice.
459 * the z loop: add missing files to Z file. Just append them (instead of
460 prepending). So one guy prepends something from the Y file from time to
461 time and another guy appends something rather frequently. Collecting
462 pond. When Y merges into Z, things get epoch and the collecting pond
463 gets smaller. What exactly are "missing files"?
465 take note of current epoch of the alpha file, let's call it the
466 recent-ts
468 find all files on disk
470 remove all files registered in the recentworld up to recent-ts
472 remove all files that have been deleted after recent-ts according to
473 recentworld
475 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
477 * rersyncrecent might be a cronjob with a (locked) state file which
478 contains things like after and maybe last z sync or such?
480 rrr-mirror might be an alternative name but how would we justify the
481 three Rs when there is no Re-Rsync-Recent?
483 With the --loop parameter it is an endless loop, without it is no loop.
484 At least this is simple.
486 * todo: new accssor z-interval specifies how often the Z file is updated
487 against the filesystem. We probably want no epoch stamp on these
488 entries. And we want to be able to filter the entries (e.g. no
489 by-modules and by-category tree)
491 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
493 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
494 into Z. We do not need the other files filled up with filesystem
495 contents though.
497 * need interface to query for a file in order to NOT call update on
498 PAUSE a second time within a short time.
500 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
502 * recommended update interval? Makes no sense, is different for
503 different users.
505 * Moosify
507 Local Variables:
508 mode: change-log
509 change-log-default-name: "Todo"
510 tab-width: 2
511 left-margin: 2
512 End: