1 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * how, exactly, do we have to deal with deletes? With rsync errors?
5 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
6 authors) failed: No such file or directory (2)
8 The file above is a delete in 1h and a new in file 1M and the
9 delete in the locally running rmirror did not get propagated to the 1M
10 object. Bug. And the consequence is a standstill.
12 It seems that a slave that works with a file below the principal needs
13 to merge things all the way up to get rid of later deletes. Or keep
14 track of all deletes and skip them later. So we need a trackdeletes.pm
15 similar to the done.pm?
17 * consider the effect when resyncing the recentfile takes longer than
18 the time per loop. Then we never rsync any file. We need to diagnose
19 that and force an increase of that loop time. But when we later are fast
20 enough again because the net has recovered, then we need to switch back
21 to original parameters.
23 * remember to verify that no temp files are left lying around and the
26 * remember: only the principal recentfile needs expiration, all others
27 shall be expired by principal if it discovers that something has move
30 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
32 * Always check if we stringify to a higher value than in the entry
35 * And in covered make an additional check if we would be able to see a
36 numerical difference between the two numbers and if we can't then switch
37 to a different, more expensive algorithm. Do not want to be caught by
40 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
42 * chaching has several aspects here: we can cache the interval of the
43 recentfile which only will change when the mtime of the file changes. We
44 must re-mirror the recentfile when its ttl has expired. Does have_read
45 tell you anything? It counts nothing at all. Only the mtime is
46 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
47 thing the have_mirrored because it is unrelated to the mtime.
49 * Robustness of floating point calculations! I always thought that the
50 string calculated by the origin server for the floating representation
51 of the epoch time is just a string. When we convert it to a number and
52 later back to a string, the other computer might come to a different
53 conclusion. This must not happen, we want to preserve it unter any
54 circumstances. I will have to write tests with overlong sequences that
55 get lost in arithmetic and must see if all still works well.
57 But one fragile point remains: if one host considers a>b and the other
58 one considers them == but no eq. To prevent this, we must probably do
61 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
63 * the concept of tracking DONE needs an object per recentfile that has
64 something like these methods:
66 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
68 covered() register() covered()
70 The unclear thing is how we translate points in time into intervals. We
71 could pass a reference to the current recent_events array when running
72 we_have(xxx) and let the DONE object iterate over it such that it only
73 has to store a list of intervals that can melt into each other. Ah, even
74 passing the list together with a list of indexes seems feasiable.
76 Or maybe ask for the inverted list?
78 Whenever the complete array is covered by the interval we say we are
79 fully covered and if the recentfile is not expired, we are uptodate.
81 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
83 * idea: have a new flag on recentfiles with the meaning: if this
84 changes, you're required to run a full rsync over all the files. The
85 reason why we set it would probably be: some foul happened. we injected
86 files in arbitrary places or didn't inject them although they changed.
87 The content of the flag? Timestamp? The relation between the
88 recentfiles would have to be inheritance from the principal, because any
89 out of band changes would soon later propagate to the next recentfile.
91 By upping the flag often one can easily ruin the slaves.
93 last out of band change? dirtymark?
95 Anyway, this implies that we read a potentially existing recentfile
98 And it implies that we have an eventloop that keeps us busy in 2-3
99 cycles, one for current stuff (tight loop) and one for the recentfiles
100 (cascade when principal has changed), one for the old stuff after a
103 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
105 * need a way to "return" the next entry after the end of a list. When
106 the caller says "before" or "after" we would like to know if he could
107 cover that interval/threshold or not because this influences the effect
108 of a newer timestamp of that recentfile. DONE with $opt{info}.
110 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
112 * one of the next things to tackle: the equivalent of csync2 -TIXU.
114 loop implies tixu (?). Nope, something like --statefile decides. Per
117 T test, I init, X including removals, U nodirtymark
119 So we have no concept of dirtymarks, we only trust that since we are
120 running we have observed everything steadily. But people will not let
121 this program run forever so we must consider both startup penalty and
122 book keeping for later runs. We keep this for later. For now we write a
123 long running mirror that merges several intervals.
125 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
127 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
130 * rersyncrecent, the script: default to one week. The name of the switch
131 is --after. Other switches? --loop!
133 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
135 * need a switch --skip-deletes (?)
137 * need a switch --enduser that tells us that the whole tempfile
138 discipline is not needed when there is no downstream user. (?)
140 Without this switch we cannot have a reasonable recent.pl that just
141 displays the recent additions. Either we accept to download everything.
142 Or we download temporary files without the typical rsync protocol
145 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
146 File::Temp::tempdir, this might be a win.
148 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
150 * apropos missing: we have no push, we never know the downstream
151 servers. People who know their downstream hosts and want to ascertain
152 something will want additional methods we have never thought about, like
153 update or delete a certain file.
155 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
157 * tempted to refactor rmirror into resolve_symlink, localize, etc.
158 Curious if rsync_options=links equal 0 vs. 1 will make the expected
161 * rsync options: it's a bit of a pain that we usually need several rsync
162 options, like compress, links, times, checksum and that there is no
163 reasonable default except the original rsync default. I think wee can
164 safely assume that the rsync options are shared between all recentfile
165 instances within one recent tree.
167 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
169 * deletes: if a delete follows an add quickly enough it may happen that
170 a downstream mirror did not see the add at all! It seems this needs to
171 be mentioned somewhere. The point here is that even if the downstream is
172 never missing the principal timeframe it may encounter a "delete" that
173 has no complimentary "add" anywhere.
175 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
177 * I suspect the treat of metadata is incorrect during read or something.
178 The bug that I am watching is that between 06:08 and 06:09 the 6h file
179 contained more than 6 hours worth of data. At 06:08 we merged into the
180 1d file. We need to take snapshots of the 6h file over the course of an
181 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
184 Much worse: watching the 1h file: right at the moment (at 06:35) it
185 covers 1218867584-1219120397 which is 70 hours.
187 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
188 08:19:44 2008, that is when I checked out last time, so it seems to be
189 aggregating and never truncating?
191 No, correct is: it is never truncating; but wrong is: it is aggregating.
192 It does receive a lot of events from time to time from a larger file.
193 Somehow a large file gets merged into the small one and because the
194 "meta/merged" attribute is missing, nobody is paying attention. I
195 believe that I can fix this by making sure that metadata are honoured
196 during read. DONE and test adjusted.
198 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
200 * grand renaming plan
202 remotebase => remoteroot to fit well with localroot DONE
203 local_path() => localroot seems to me should already work DONE
204 recentfile_basename => rfilename no need to stress it has no slash DONE
206 filenameroot??? Doesn't seem too bad to me today. Maybe something like
207 kern? It would anyway need a deprecation cycle because it is an
208 important constructor.
210 * I like the portability that Data::Serializer brings us but the price
211 is that some day we might find out that it is slowing us a bit. We'll
214 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
216 * should we not enter the interval of the principal (or the interval of
217 the merging file?) in every aggregated/merged file?
219 * we should aim at a first release and give up on thinking about
220 sanitizing stuff and zloop. Let's just admit that a full traditional
221 rsync is the only available sanitizer ATM. Otherwise it's complicated
222 stuff: sanitizing on the origin server, sanitizing on the slaves,
223 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
224 get the basics out before this becomes a major cause for mess.
226 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
228 * On OSes not supporting symlinks we expect that RECENT.recent contains
229 the contents of the principal recentfile. Actually this is identical on
230 systems supporting symlinks. Simple, what follows from that is that we
231 need to keep the serializer in the metadata because we cannot read it
232 from the filename, doesn't it? Of course not. It's a chicken and egg
233 problem. This leaves us with the problem to actually parse the
234 serialized data to find out in which format it is. So who can do the 4
235 or 5 magics we wanted to support? File::LibMagic?
237 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
239 * remotebase and recentfile_basename are ugly names. Now that we need a
240 word for the shortest/principal/driving recentfile too we should do
243 localroot is good. rfile is good. local_path() is bad, local_path($path)
244 is medium, filenameroot() is bad, remotebase is bad, recentfile is
247 Up to now remotebase was the string that described the remote root
248 directory in rsync notation, like pause.perl.org::authors. And
249 recentfile_basename was "RECENT-1h.yaml".
251 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
253 * The test that was added in today's checkin is a good start for a test
254 of rmirror. We should have more methods in Recent.pm: verify,
255 addmissingfiles. We should verify the current tree, then rmirror it and
256 then verifytree the copy. We could then add some arbitrary file and let
257 it be discovered by addmissingfiles, then rmirror again and then
258 verifytree the copy again.
260 Then we could start stealing from csync2 sqlite database [no port to
261 OSX!] and fill a local DB. And methods to compare the database with the
262 recentfiles. Our strength is that in principle we could maintain state
263 with a single float. We have synced up to 1234567890.123456. If the Z
264 file does not add new files all we have to do is mirror the new ones and
267 This makes it clear that we should extend current protocol and declare
268 that we cheat when we add files too late, just to help the other end
269 keeping track. Ah yes, that's what was meant when zloop was mentioned
272 Maybe need to revisit File::Mirror to help me with this task.
274 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
276 * There must be an allow-me-to-truncate flag in every recentfile.
277 Without it one could construct a sequence of updates winning the locking
278 battle against the aggregator. Only if an aggregator has managed to
279 merge data over to the next level, truncating can be allowed. DONE with
282 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
284 * We should probably guarantee that no duplicates enter the aggregator
287 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
289 * To get merge operation faster would need a good benchmark test. What
290 02 spits out isn't reliable enough and is dominated by many other
293 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
294 Author: Andreas J. Koenig <k@k75.(none)>
295 Date: Sat Aug 2 07:58:04 2008 +0200
299 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
300 Author: Andreas J. Koenig <k@k75.(none)>
301 Date: Sat Aug 2 11:40:29 2008 +0200
303 there was a lot of trying but the effect is hardly measurable with
306 * overhead of connecting seems high. When setting
307 max_files_per_connection to 1 we see that.
309 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
311 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
313 25138.2215600014/3600 = 6.98283932222261
315 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
316 be a bug in the aggregator. FIXED
318 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
320 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
321 we have it, can we mirror it, mirror it!
323 I fear this needs a new class which might be called
324 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
325 and treat them as an entity. I realize that a single recentfile may be
326 sufficient for certain tasks and that it is handy for the low level
327 programmer but it is not nice to use. If there is a delete in the 1h
328 file then the 6h file still contains it. Seekers of the best information
329 need to combine at least some of the recentfiles most of the time.
331 There is the place for the Z loop!
333 But the combination is something to collect in a database, isn't it. Did
334 csync2 just harrumph?
336 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
338 * it just occurred to me that hosts in the same mirroring pool could
339 help out each other even without rewriting the recentfile. Just fetch
340 the stuff to mirror from several places, bingo. But that's something
341 that should rather live in a separate package or in rsync directly.
343 * cronjobs are unsuited because with ntp they would all come at the full
344 minute and disturb each other. Besides that I'd hate to have a backbone
345 with more than a few seconds latency.
347 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
349 * a second rsync server with access control for PAUSE. Port? 873 is the
350 standard port, let's take 8873.
352 * if there were a filesystem based on this, it would have a slow access
353 to inexistent files. It would probably provide wrong readdir (only based
354 on current content) or also a slow one (based on a recentfile written
355 after the call). But it would provide fast access to existing files. Or
356 one would deliberately allow slightly blurred answers based on some
357 sqlite reflection of the recentfiles.
359 * todo: write a variant of mirror() that combines two or more
360 recentfiles and treats them like one
362 * todo: signal handler to remove the tempfile
364 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
366 * now that we have the symlink I forgot how it should be used in
369 * the z loop: add missing files to Z file. Just append them (instead of
370 prepending). So one guy prepends something from the Y file from time to
371 time and another guy appends something rather frequently. Collecting
372 pond. When Y merges into Z, things get epoch and the collecting pond
373 gets smaller. What exactly are "missing files"?
375 take note of current epoch of the alpha file, let's call it the
378 find all files on disk
380 remove all files registered in the recentworld up to recent-ts
382 remove all files that have been deleted after recent-ts according to
385 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
387 * rersyncrecent might be a cronjob with a (locked) state file which
388 contains things like after and maybe last z sync or such?
390 rrr-mirror might be an alternative name but how would we justify the
391 three Rs when there is no Re-Rsync-Recent?
393 With the --loop parameter it is an endless loop, without it is no loop.
394 At least this is simple.
396 * todo: new accssor z-interval specifies how often the Z file is updated
397 against the filesystem. We probably want no epoch stamp on these
398 entries. And we want to be able to filter the entries (e.g. no
399 by-modules and by-category tree)
401 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
403 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
404 into Z. We do not need the other files filled up with filesystem
407 * need interface to query for a file in order to NOT call update on
408 PAUSE a second time within a short time.
410 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
412 * recommended update interval? Makes no sense, is different for
419 change-log-default-name: "Todo"