| blob | 706ec6348db20e08e3fec9606668c7c43559921e |
1 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
4 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
6 * statusfile, maybe only on demand, alone to have a sharp debugging
7 tool. It is locked and all recentfiles dump themselves into it and we
8 can build a viewer that lets us know where we stand and what's inside.
10 * how, exactly, do we have to deal with deletes? With rsync errors?
12 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
13 authors) failed: No such file or directory (2)
15 The file above is a delete in 1h and a new in file 1M and the
16 delete in the locally running rmirror did not get propagated to the 1M
17 object. Bug. And the consequence is a standstill.
19 It seems that a slave that works with a file below the principal needs
20 to merge things all the way up to get rid of later deletes. Or keep
21 track of all deletes and skip them later. So we need a trackdeletes.pm
22 similar to the done.pm?
24 * consider the effect when resyncing the recentfile takes longer than
25 the time per loop. Then we never rsync any file. We need to diagnose
26 that and force an increase of that loop time. But when we later are fast
27 enough again because the net has recovered, then we need to switch back
28 to original parameters.
30 * remember to verify that no temp files are left lying around and the
31 signal handler
33 * remember: only the principal recentfile needs expiration, all others
34 shall be expired by principal if it discovers that something has move
35 upstream.
37 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
39 * Always check if we stringify to a higher value than in the entry
40 before.
42 * And in covered make an additional check if we would be able to see a
43 numerical difference between the two numbers and if we can't then switch
44 to a different, more expensive algorithm. Do not want to be caught by
45 floating surprises.
47 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
49 * chaching has several aspects here: we can cache the interval of the
50 recentfile which only will change when the mtime of the file changes. We
51 must re-mirror the recentfile when its ttl has expired. Does have_read
52 tell you anything? It counts nothing at all. Only the mtime is
53 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
54 thing the have_mirrored because it is unrelated to the mtime.
56 * Robustness of floating point calculations! I always thought that the
57 string calculated by the origin server for the floating representation
58 of the epoch time is just a string. When we convert it to a number and
59 later back to a string, the other computer might come to a different
60 conclusion. This must not happen, we want to preserve it unter any
61 circumstances. I will have to write tests with overlong sequences that
62 get lost in arithmetic and must see if all still works well.
64 But one fragile point remains: if one host considers a>b and the other
65 one considers them == but no eq. To prevent this, we must probably do
66 some extra homework.
68 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
70 * the concept of tracking DONE needs an object per recentfile that has
71 something like these methods:
73 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
75 covered() register() covered()
77 The unclear thing is how we translate points in time into intervals. We
78 could pass a reference to the current recent_events array when running
79 we_have(xxx) and let the DONE object iterate over it such that it only
80 has to store a list of intervals that can melt into each other. Ah, even
81 passing the list together with a list of indexes seems feasiable.
83 Or maybe ask for the inverted list?
85 Whenever the complete array is covered by the interval we say we are
86 fully covered and if the recentfile is not expired, we are uptodate.
88 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
90 * idea: have a new flag on recentfiles with the meaning: if this
91 changes, you're required to run a full rsync over all the files. The
92 reason why we set it would probably be: some foul happened. we injected
93 files in arbitrary places or didn't inject them although they changed.
94 The content of the flag? Timestamp? The relation between the
95 recentfiles would have to be inheritance from the principal, because any
96 out of band changes would soon later propagate to the next recentfile.
98 By upping the flag often one can easily ruin the slaves.
100 last out of band change? dirtymark?
102 Anyway, this implies that we read a potentially existing recentfile
103 before we write one.
105 And it implies that we have an eventloop that keeps us busy in 2-3
106 cycles, one for current stuff (tight loop) and one for the recentfiles
107 (cascade when principal has changed), one for the old stuff after a
108 dirtymark change.
110 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
112 * need a way to "return" the next entry after the end of a list. When
113 the caller says "before" or "after" we would like to know if he could
114 cover that interval/threshold or not because this influences the effect
115 of a newer timestamp of that recentfile. DONE with $opt{info}.
117 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
119 * one of the next things to tackle: the equivalent of csync2 -TIXU.
121 loop implies tixu (?). Nope, something like --statefile decides. Per
122 default we do ...?
124 T test, I init, X including removals, U nodirtymark
126 So we have no concept of dirtymarks, we only trust that since we are
127 running we have observed everything steadily. But people will not let
128 this program run forever so we must consider both startup penalty and
129 book keeping for later runs. We keep this for later. For now we write a
130 long running mirror that merges several intervals.
132 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
134 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
135 the intervals!
137 * rersyncrecent, the script: default to one week. The name of the switch
138 is --after. Other switches? --loop!
140 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
142 * need a switch --skip-deletes (?)
144 * need a switch --enduser that tells us that the whole tempfile
145 discipline is not needed when there is no downstream user. (?)
147 Without this switch we cannot have a reasonable recent.pl that just
148 displays the recent additions. Either we accept to download everything.
149 Or we download temporary files without the typical rsync protocol
150 advantages.
152 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
153 File::Temp::tempdir, this might be a win.
155 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
157 * apropos missing: we have no push, we never know the downstream
158 servers. People who know their downstream hosts and want to ascertain
159 something will want additional methods we have never thought about, like
160 update or delete a certain file.
162 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
164 * tempted to refactor rmirror into resolve_symlink, localize, etc.
165 Curious if rsync_options=links equal 0 vs. 1 will make the expected
166 difference.
168 * rsync options: it's a bit of a pain that we usually need several rsync
169 options, like compress, links, times, checksum and that there is no
170 reasonable default except the original rsync default. I think wee can
171 safely assume that the rsync options are shared between all recentfile
172 instances within one recent tree.
174 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
176 * deletes: if a delete follows an add quickly enough it may happen that
177 a downstream mirror did not see the add at all! It seems this needs to
178 be mentioned somewhere. The point here is that even if the downstream is
179 never missing the principal timeframe it may encounter a "delete" that
180 has no complimentary "add" anywhere.
182 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
184 * I suspect the treat of metadata is incorrect during read or something.
185 The bug that I am watching is that between 06:08 and 06:09 the 6h file
186 contained more than 6 hours worth of data. At 06:08 we merged into the
187 1d file. We need to take snapshots of the 6h file over the course of an
188 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
189 enough.
191 Much worse: watching the 1h file: right at the moment (at 06:35) it
192 covers 1218867584-1219120397 which is 70 hours.
194 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
195 08:19:44 2008, that is when I checked out last time, so it seems to be
196 aggregating and never truncating?
198 No, correct is: it is never truncating; but wrong is: it is aggregating.
199 It does receive a lot of events from time to time from a larger file.
200 Somehow a large file gets merged into the small one and because the
201 "meta/merged" attribute is missing, nobody is paying attention. I
202 believe that I can fix this by making sure that metadata are honoured
203 during read. DONE and test adjusted.
205 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
207 * grand renaming plan
209 remotebase => remoteroot to fit well with localroot DONE
210 local_path() => localroot seems to me should already work DONE
211 recentfile_basename => rfilename no need to stress it has no slash DONE
213 filenameroot??? Doesn't seem too bad to me today. Maybe something like
214 kern? It would anyway need a deprecation cycle because it is an
215 important constructor.
217 * I like the portability that Data::Serializer brings us but the price
218 is that some day we might find out that it is slowing us a bit. We'll
219 see.
221 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
223 * should we not enter the interval of the principal (or the interval of
224 the merging file?) in every aggregated/merged file?
226 * we should aim at a first release and give up on thinking about
227 sanitizing stuff and zloop. Let's just admit that a full traditional
228 rsync is the only available sanitizer ATM. Otherwise it's complicated
229 stuff: sanitizing on the origin server, sanitizing on the slaves,
230 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
231 get the basics out before this becomes a major cause for mess.
233 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
235 * On OSes not supporting symlinks we expect that RECENT.recent contains
236 the contents of the principal recentfile. Actually this is identical on
237 systems supporting symlinks. Simple, what follows from that is that we
238 need to keep the serializer in the metadata because we cannot read it
239 from the filename, doesn't it? Of course not. It's a chicken and egg
240 problem. This leaves us with the problem to actually parse the
241 serialized data to find out in which format it is. So who can do the 4
242 or 5 magics we wanted to support? File::LibMagic?
244 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
246 * remotebase and recentfile_basename are ugly names. Now that we need a
247 word for the shortest/principal/driving recentfile too we should do
248 something about it.
250 localroot is good. rfile is good. local_path() is bad, local_path($path)
251 is medium, filenameroot() is bad, remotebase is bad, recentfile is
252 already deprecated.
254 Up to now remotebase was the string that described the remote root
255 directory in rsync notation, like pause.perl.org::authors. And
256 recentfile_basename was "RECENT-1h.yaml".
258 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
260 * The test that was added in today's checkin is a good start for a test
261 of rmirror. We should have more methods in Recent.pm: verify,
262 addmissingfiles. We should verify the current tree, then rmirror it and
263 then verifytree the copy. We could then add some arbitrary file and let
264 it be discovered by addmissingfiles, then rmirror again and then
265 verifytree the copy again.
267 Then we could start stealing from csync2 sqlite database [no port to
268 OSX!] and fill a local DB. And methods to compare the database with the
269 recentfiles. Our strength is that in principle we could maintain state
270 with a single float. We have synced up to 1234567890.123456. If the Z
271 file does not add new files all we have to do is mirror the new ones and
272 delete the goners.
274 This makes it clear that we should extend current protocol and declare
275 that we cheat when we add files too late, just to help the other end
276 keeping track. Ah yes, that's what was meant when zloop was mentioned
277 earlier.
279 Maybe need to revisit File::Mirror to help me with this task.
281 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
283 * There must be an allow-me-to-truncate flag in every recentfile.
284 Without it one could construct a sequence of updates winning the locking
285 battle against the aggregator. Only if an aggregator has managed to
286 merge data over to the next level, truncating can be allowed. DONE with
287 accessor merged.
289 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
291 * We should probably guarantee that no duplicates enter the aggregator
292 array.
294 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
296 * To get merge operation faster would need a good benchmark test. What
297 02 spits out isn't reliable enough and is dominated by many other
298 things. Between
300 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
301 Author: Andreas J. Koenig <k@k75.(none)>
302 Date: Sat Aug 2 07:58:04 2008 +0200
304 and
306 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
307 Author: Andreas J. Koenig <k@k75.(none)>
308 Date: Sat Aug 2 11:40:29 2008 +0200
310 there was a lot of trying but the effect is hardly measurable with
311 current tests.
313 * overhead of connecting seems high. When setting
314 max_files_per_connection to 1 we see that.
316 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
318 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
320 25138.2215600014/3600 = 6.98283932222261
322 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
323 be a bug in the aggregator. FIXED
325 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
327 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
328 we have it, can we mirror it, mirror it!
330 I fear this needs a new class which might be called
331 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
332 and treat them as an entity. I realize that a single recentfile may be
333 sufficient for certain tasks and that it is handy for the low level
334 programmer but it is not nice to use. If there is a delete in the 1h
335 file then the 6h file still contains it. Seekers of the best information
336 need to combine at least some of the recentfiles most of the time.
338 There is the place for the Z loop!
340 But the combination is something to collect in a database, isn't it. Did
341 csync2 just harrumph?
343 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
345 * it just occurred to me that hosts in the same mirroring pool could
346 help out each other even without rewriting the recentfile. Just fetch
347 the stuff to mirror from several places, bingo. But that's something
348 that should rather live in a separate package or in rsync directly.
350 * cronjobs are unsuited because with ntp they would all come at the full
351 minute and disturb each other. Besides that I'd hate to have a backbone
352 with more than a few seconds latency.
354 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
356 * a second rsync server with access control for PAUSE. Port? 873 is the
357 standard port, let's take 8873.
359 * if there were a filesystem based on this, it would have a slow access
360 to inexistent files. It would probably provide wrong readdir (only based
361 on current content) or also a slow one (based on a recentfile written
362 after the call). But it would provide fast access to existing files. Or
363 one would deliberately allow slightly blurred answers based on some
364 sqlite reflection of the recentfiles.
366 * todo: write a variant of mirror() that combines two or more
367 recentfiles and treats them like one
369 * todo: signal handler to remove the tempfile
371 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
373 * now that we have the symlink I forgot how it should be used in
374 practice.
376 * the z loop: add missing files to Z file. Just append them (instead of
377 prepending). So one guy prepends something from the Y file from time to
378 time and another guy appends something rather frequently. Collecting
379 pond. When Y merges into Z, things get epoch and the collecting pond
380 gets smaller. What exactly are "missing files"?
382 take note of current epoch of the alpha file, let's call it the
383 recent-ts
385 find all files on disk
387 remove all files registered in the recentworld up to recent-ts
389 remove all files that have been deleted after recent-ts according to
390 recentworld
392 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
394 * rersyncrecent might be a cronjob with a (locked) state file which
395 contains things like after and maybe last z sync or such?
397 rrr-mirror might be an alternative name but how would we justify the
398 three Rs when there is no Re-Rsync-Recent?
400 With the --loop parameter it is an endless loop, without it is no loop.
401 At least this is simple.
403 * todo: new accssor z-interval specifies how often the Z file is updated
404 against the filesystem. We probably want no epoch stamp on these
405 entries. And we want to be able to filter the entries (e.g. no
406 by-modules and by-category tree)
408 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
410 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
411 into Z. We do not need the other files filled up with filesystem
412 contents though.
414 * need interface to query for a file in order to NOT call update on
415 PAUSE a second time within a short time.
417 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
419 * recommended update interval? Makes no sense, is different for
420 different users.
422 * Moosify
424 Local Variables:
425 mode: change-log
426 change-log-default-name: "Todo"
427 tab-width: 2
428 left-margin: 2
429 End: