| blob | 4c52c7ca03bfdf508278b560a11ae7acfb5d7a71 |
1 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * another bug is the fact that the mirror command deletes files before
4 it unhides the index file, thus confusing downstream slaves. We must not
5 delete before unhiding and must delete after unhiding. FIXED.
7 * new complication about the slave that is playing a server role.
8 Currently we mirror from newest to oldest with a hidden temporary file
9 as index. And when one file is finished, we unhide the index file.
10 Imagine the cascading server/slave is dead for a day. It then starts
11 mirroring again with the freshest thing and unhides the freshest index
12 file when it has worked through it. In that moment it exposes a time
13 hole. Because it now works on the second recentfile which is still
14 hidden.
16 We currently do nothing special to converge after such a drop out. At
17 least not intentionally and robustly and thought through.
19 The algorithm we use to seed the next file needs quite a lot of more
20 robustness than it currently has. Something to do with looking at the
21 merged element of the next rf and when it has dropped off, we seed
22 immediately. And if it ramains dropped off, we seed again, of course.
24 Nope, looking from smaller to larger RFS we look at the merged element
25 of this RF and at the minmax/max element of the next RF. If that
26 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
27 seeding it.
29 And we need a public accessor seed and unseed or seeded. But not the mix
30 of public and private stuff that then is used behind the back.
32 And then the secondary* stuff must go.
34 And we must understand what the impact is on the DONE system. Can it go
35 unnoticed that there was a hole? And could the DONE system have decided
36 the hole is covered? This sould be testable with three directories where
37 the middle stops working for a while. Done->merge is suspicious, we must
38 stop it from merging non-conflatable neighbors due to broken continuity.
40 * dirtymark now lives in Recentfile, needs to be used in rmirror.
42 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
44 * Slaven suggests to have the current epoch or the whole current
45 recentfile available from the HTTP server and take it away with
46 keepalive. This direction goes the granularity down to subseconds.
48 We might want to rewrite everything to factor out transport and allow
49 the whole thing to run via HTTP.
51 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
53 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
54 symlink?
56 * smoker on k81 fetching from k75 to verify cascading works. See
57 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
59 * maybe the loop should wait for CHECKSUMS file after every upload. And
60 CPAN.pm needs to deal with timestamps in the future.
62 * do not forget the dirtymark!
64 Text: have a new flag on recentfiles with the meaning: if this
65 changes, you're required to run a full rsync over all the files. The
66 reason why we set it would probably be: some foul happened. we injected
67 files in arbitrary places or didn't inject them although they changed.
68 The content of the flag? Timestamp? The relation between the
69 recentfiles would have to be inheritance from the principal, because any
70 out of band changes would soon later propagate to the next recentfile.
72 By upping the flag often one can easily ruin the slaves.
74 last out of band change? dirtymark?
76 Anyway, this implies that we read a potentially existing recentfile
77 before we write one.
79 And it implies that we have an eventloop that keeps us busy in 2-3
80 cycles, one for current stuff (tight loop) and one for the recentfiles
81 (cascade when principal has changed), one for the old stuff after a
82 dirtymark change.
84 And it implies that the out-of-band change in any of the recentfiles
85 must have a lock on the principal file and there is the place to set the
86 dirtymark.
88 * after the dirtymark is done: fill up recentfiles with fake (historic)
89 entries; fill up with individual corrections; algorithm maybe to be done
90 with bigfloat so that we can always place something in the middle
91 between two entries. Before we must switch to bigfloat we could try to
92 use Data::Float::nextup to get the.
94 * lookup by epoch and by path and use this ability on the pause to never
95 again register a file twice that doesn't need it.
97 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
98 directly.
100 * start a FAQ, especially quick start guide questions. Also to aid those
101 problematic areas where we have no good solution, like the "links"
102 option to rsync.
104 * wish feedback when we are slow.
106 * reduce mccabe
108 * Remove a few DEBUG statements.
110 * The multiple-rrr way of doing things needs a new option to rmirror,
111 like piecemeal or so. Not urgent because after the first pass through,
112 things run smoothely. It's only ugly during the first pass.
114 * I have the suspicion that the code is broken that decides if the
115 neighboring RF needs to be seeded. I fear when too much time has gone
116 between two calls (in our case more than one hour), it would not seed
117 the neighbor. Of course this will never be noticed, so we need a good
118 test for it.
120 * local/localroot confusion: I currently pass both options but one must
121 do.
123 * accounts for early birds on PAUSE rsync daemon.
125 * hardcoded 20 seconds
127 * who mirrors the index? DOING now.
129 * which CPAN mirrors offer rsync?
131 * visit all XXX, visit all _float places
133 * rename the pathdb stuff, it's too confusing. No idea how.
135 * rrr-inotify, backpan, rrr-register
137 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
139 * current bugs: the pathdb seems to get no reset, the seeding of the
140 secondaryttl stuff seems not to have an effect. Have helped myself with
141 a rand(10), need to fix this back. So not checked in. Does the rand
142 thing even help?
144 The rand thing helps. The secondaryttl stuff was in the wrong line,
145 fixed now.
147 The pathdb stuff was because I called either _pathdb or __pathdb on the
148 wrong object. FIXED now.
150 * It's not so beautiful if we never fetch the recentfiles that are not
151 the principal, even if this is correct behaviour. We really do not need
152 them after we have fetched the whole content.
154 OK, we want a switch for that: secondaryttl DONE
156 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
158 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
159 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
160 it lets 40 follow. FIXED
162 * See that the long running process really only updates the principal
163 file unless it has missed a timespan during which something happened. If
164 nothing happened, it must notice even when it misses the timespan. DONE
166 * we must throw away the pathdb when we have reached the end of Z. From
167 that moment we can have a very small pathdb because the only reason for
168 a pathdb is that we know to ignore old records in old files. We won't
169 need this pathdb again before the next full pass over the data is
170 necessary and then we will rebuild it as we go along. DONE
172 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
174 * I think, Done::register_one is doing wrong in that it does not
175 conflate neighboring pieces. The covered() method cannot do this because
176 it has no recent_events array at hand. But register_one has it and could
177 do it and for some reason misses to do it (sometimes).
179 This means that the three tests I just wrote can probably not survive
180 because they test with an already broken Done structure.
182 The art now is to detect how it happens, then to reproduce, then write a
183 test, then fix it.
185 So from the logfile this is what happens: we have a good interval with
186 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
187 top of it. Locally we now mirror F2 and open a new done interval for it.
188 Then we mirror F1 but this time with the timestamp T1b. And when we then
189 try to close the gap, we do not find T1 but instead something older. We
190 should gladly accept this older piece and this would fix this bug.
192 FIXED
194 * bug to fix: when the 1h file changes while rmirror is running, we do
195 correctly sync the new files but never switch to the 6h file but rather
196 stay in a rather quick loop that fetches the 1h file again and again.
198 Is it possible that we initialize a new object? Or does
199 get_remote_recentfile_as_tempfile overwrite something in myself?
201 Want a new option: _runstatusfile => $file which frequently dumps the
202 state of all recentfiles to a file.
204 FIXED
206 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
208 * Todo: now teach update to verify the timestamp is about to write
209 against the previous and use _increase_a_bit if it doesn't comply with
210 strict monotony. DONE
212 * The problem of rounding. So far perl's default precision was
213 sufficient. One day it won't be. FakeFloat has an easy job when it is
214 only reading and other machines have written correctly. But when we want
215 to write a floating point number that is a bit larger than the other
216 one, then we need our own idea of precision.
218 Slaven said: just append a "1". This might be going towards the end of
219 usability too quickly. I'd like something that actually uses the decimal
220 system. Well, appending a 1 also does this but...
222 E.g. we have 1.0. nextup on this architecture is starting with
223 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
224 taken the 1.0000000000000003 and the next user comes and the time tells
225 him 1.0 again. He has to beat my number without stepping over the
226 nextup. This is much less space than I had when I chose 1,2,3.
228 What is also irritating is that nextup is architecture dependent. The
229 128 bit guy must choose very long numbers to fit in between whereas the
230 other one with 16 bit uses larger steps. But then the algorithm is the
231 same for both, so that would be a nice thing.
233 I see two situation where we need this. One is when Time::HiRes returns
234 us a value that is <= the last entry in our recentfile. In this case
235 (let's call it the end-case) we must fill the region between that number
236 and the next higher native floating point number. The other is when we
237 inject an old file into an old recentfile (we would then also set a new
238 dirtymark). We find the integer value already taken and need a slightly
239 different one (let's call it the middle-case). The difference between
240 the two situations is that the next user will want to find something
241 higher than my number in the end-case and something lower than my number
242 in the middle case.
244 So I suggest we give the function both a value and an upper bound and it
245 calculates us a primitive middle. The upper bound in the middle-case is
246 the next integer. The upper bound on the end-case is the nextup floating
247 point number. But the latter poses another problem: if we have occupied
248 the middle m between x and nextup(x), then the nextup(m) will probably
249 not be the same as nextup(x) because some rounding will take place
250 before the nextup is calculated and when the rounding reaches the
251 nextup(x), we will end up at nextup(nextup(x)).
253 So we really need to consider the nextup and the nextdown from there and
254 then the middle and that's the number we may approach asymptotically.
255 Ugly. But DONE.
257 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
259 * consider deprecating the use of RECENT.recent as a symlink. It turns
260 out to need extra hoops with the rsync options and just isn't worth it.
261 Or maybe these extra hoops are needed anyway for the rest of the tree?
262 Nope, can't be the case because not all filesystems support symlinks.
264 But before doing the large step, I'll deprecate the call of
265 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
266 only introduced to resolve RECENT.recent and complicates the routine far
267 beyond what it deserves.
269 DONE. Won't deprecate RECENT.recent, just moved its handling to the
270 supervisor.
272 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
274 * I think it's a bug that the rsync_option links must be set to true in
275 order to support RECENT.recent and that nobody cares to set it
276 automatically. Similar for ignore_link_stat_errors.
278 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
280 * Todo: find all todos together and make a plan what is missing for a
281 release.
283 - verifytree or something like that. fsck maybe.
285 - rersyncrecent, the script itself? What it do?
287 - a way to only mirror the recentfiles without mirroring the whole
288 remote system such that people can decide to mirror only partially see
289 also 2008-08-30. .shadow-xxx directory? this also needed for a
290 filesystem that is still incomplete and might need the mirrorfiles for
291 lookup(?)
293 - long living objects that mirror again and again. Inject something
294 into ta, see how it goes over to tb.
296 - how do we continue filling up the DONE system when we use an object
297 for the second time? "fully covered" and "uptodate" or new terminology.
299 - overview called on the wrong file should be understandable
301 - the meta data field that must change when we fake something up so that
302 the downstream people know they have to re-fetch everything.
304 - how tolerant are we against missing files upstream? how do we keep
305 track? there are legitimate cases where we did read upstream index right
306 before a file got deleted there and then find that file as new and want
307 it. There are other cases that are not self healing and must be tracked
308 and bugreported.
310 - how, exactly, do we have to deal with deletes? With rsync errors?
312 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
313 authors) failed: No such file or directory (2)
315 The file above is a delete in 1h and a new in file 1M and the
316 delete in the locally running rmirror did not get propagated to the 1M
317 object. Bug. And the consequence is a standstill.
319 It seems that a slave that works with a file below the principal needs
320 to merge things all the way up to get rid of later deletes. Or keep
321 track of all deletes and skip them later. So we need a trackdeletes.pm
322 similar to the done.pm?
324 see also 2008-08-20 about spurious deletes that really have no add
325 counterpart and yet they are not wrong.
327 - consider the effect when resyncing the recentfile takes longer than
328 the time per loop. Then we never rsync any file. We need to diagnose
329 that and force an increase of that loop time. But when we later are fast
330 enough again because the net has recovered, then we need to switch back
331 to original parameters. ERm, no, it's enough to keep syncing at least
332 one file before refetching an index file.
334 - remember to verify that no temp files are left lying around and the
335 signal handler
337 - status file for not long running jobs that want to track upstream with
338 a, say, cronjob.
340 - revisit all XXX _float areas and study Sub::Exporter DONE
342 - persistent DB even though we just said we do not need it. Just for
343 extended capabilities and time savings when, for example, upstream
344 announces a reset and we get new recentfiles and could then limit
345 ourselves to a subset of files (those that have a changed epoch) in a
346 first pass and would only then do the loop to verify the rest. Or
347 something.
349 * Todo: aggregate files should know their feed and finding the principal
350 should be done stepwise. (?)
352 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
353 left locks around. DONE
355 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
357 * maybe extend the _overview so that it always says if and where the
358 last file is in the next file and where the next event in the next rf
359 would lie. No, don't like this anymore. REJECT
361 * take the two new redundant tests out again, only the third must
362 survive. DONE
364 * Todo: add a sanity check if the merged structure is really pointing to
365 a different rf and that this different rf is larger. DONE
367 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
369 * now test, if they are overlapping. And test if there is a file in the
370 next rf that would fit into this rf's interval.
372 1h 1222324012.8474 1222322541.7963 0.4086
373 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
374 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
375 1W 1222313218.3626 1221708477.5829 167.9835
377 I suspect that somebody writes a merged timestamp without having merged
378 and then somebody else relies on it.
380 If aggregate is running, the intervals must not be extravagated, if it
381 is not running, there must not be bounds, the total number of events in
382 the system must be counted and must be controlled throughout the tests.
383 That the test required the additional update was probably nonsense,
384 because aggregate can cut pieces too. FIXED & DONE
386 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
388 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
389 changed. FIXED
391 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
393 * Most apparent bug at the moment is that the recentfiles are fetched
394 too often. Only the principal should be fetched and if it has not
395 changed, the others should not be refetched. ATM I must admit that I'm
396 happy that we refetch more often than needed because I can more easily
397 fix bugs while the thing is running.
399 * Let's say, 1220474966.19501 is a timestamp of a file that is already
400 done but the done system does not know about it. The reason for the
401 failure is not known and we never reach the status uptodate because of
402 this. We must get over it.
404 Later it turns out that the origin server had a bug somewhere.
405 1220474966.19042 came after 1220474966.19501. Or better: it was in the
406 array of the recentfile one position above. The bug was my own.
408 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
410 * There is the race condition where the server does a delete and the
411 slave does not yet know and then tries to download it because he sees
412 the new. So for this time window we must be more tolerant against
413 failure. If we cannot download a file, we should just skip it and should
414 not retry immediately. The whole system should discover the lost thing
415 later. Keeping track with the DONE system should really be a no brainer.
417 But there is something more: the whole filesystem is a database and the
418 recentfiles are one possible representation of it. It's a pretty useful
419 representation I think that's why I have implemented something around
420 it. But for strictly local operation it has little value. For local
421 operation we would much rather have a database. So we would enter every
422 recentfile reading and every rsync operation and for every file the last
423 state change and what it leads to. Then we would always ignore older
424 records without the efforts involved with recentfiles.
426 The database would have: path,recentepoch,rsyncedon,deletedon
428 Oh well, not yet clear where this leads to.
430 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
432 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
433 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
435 * statusfile, maybe only on demand, alone to have a sharp debugging
436 tool. It is locked and all recentfiles dump themselves into it and we
437 can build a viewer that lets us know where we stand and what's inside.
439 * remember: only the principal recentfile needs expiration, all others
440 shall be expired by principal if it discovers that something has move
441 upstream.
443 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
445 * Always check if we stringify to a higher value than in the entry
446 before. DONE
448 * And in covered make an additional check if we would be able to see a
449 numerical difference between the two numbers and if we can't then switch
450 to a different, more expensive algorithm. Do not want to be caught by
451 floating surprises. DONE
453 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
455 * caching has several aspects here: we can cache the interval of the
456 recentfile which only will change when the mtime of the file changes. We
457 must re-mirror the recentfile when its ttl has expired. Does have_read
458 tell you anything? It counts nothing at all. Only the mtime is
459 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
460 thing the have_mirrored because it is unrelated to the mtime.
462 * Robustness of floating point calculations! I always thought that the
463 string calculated by the origin server for the floating representation
464 of the epoch time is just a string. When we convert it to a number and
465 later back to a string, the other computer might come to a different
466 conclusion. This must not happen, we want to preserve it under any
467 circumstances. I will have to write tests with overlong sequences that
468 get lost in arithmetic and must see if all still works well. DONE
470 But one fragile point remains: if one host considers a>b and the other
471 one considers them == but no eq. To prevent this, we must probably do
472 some extra homework. DONE
474 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
476 * the concept of tracking DONE needs an object per recentfile that has
477 something like these methods:
479 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
481 covered() register() covered()
483 The unclear thing is how we translate points in time into intervals. We
484 could pass a reference to the current recent_events array when running
485 we_have(xxx) and let the DONE object iterate over it such that it only
486 has to store a list of intervals that can melt into each other. Ah, even
487 passing the list together with a list of indexes seems feasiable.
489 Or maybe ask for the inverted list?
491 Whenever the complete array is covered by the interval we say we are
492 fully covered and if the recentfile is not expired, we are uptodate.
494 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
496 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
498 * need a way to "return" the next entry after the end of a list. When
499 the caller says "before" or "after" we would like to know if he could
500 cover that interval/threshold or not because this influences the effect
501 of a newer timestamp of that recentfile. DONE with $opt{info}.
503 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
505 * one of the next things to tackle: the equivalent of csync2 -TIXU.
507 loop implies tixu (?). Nope, something like --statefile decides. Per
508 default we do ...?
510 T test, I init, X including removals, U nodirtymark
512 So we have no concept of dirtymarks, we only trust that since we are
513 running we have observed everything steadily. But people will not let
514 this program run forever so we must consider both startup penalty and
515 book keeping for later runs. We keep this for later. For now we write a
516 long running mirror that merges several intervals.
518 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
520 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
521 the intervals!
523 * rersyncrecent, the script: default to one week. The name of the switch
524 is --after. Other switches? --loop!
526 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
528 * need a switch --skip-deletes (?)
530 * need a switch --enduser that tells us that the whole tempfile
531 discipline is not needed when there is no downstream user. (?)
533 Without this switch we cannot have a reasonable recent.pl that just
534 displays the recent additions. Either we accept to download everything.
535 Or we download temporary files without the typical rsync protocol
536 advantages.
538 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
539 File::Temp::tempdir, this might be a win.
541 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
543 * apropos missing: we have no push, we never know the downstream
544 servers. People who know their downstream hosts and want to ascertain
545 something will want additional methods we have never thought about, like
546 update or delete a certain file.
548 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
550 * tempted to refactor rmirror into resolve_symlink, localize, etc.
551 Curious if rsync_options=links equal 0 vs. 1 will make the expected
552 difference.
554 * rsync options: it's a bit of a pain that we usually need several rsync
555 options, like compress, links, times, checksum and that there is no
556 reasonable default except the original rsync default. I think wee can
557 safely assume that the rsync options are shared between all recentfile
558 instances within one recent tree.
560 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
562 * deletes: if a delete follows an add quickly enough it may happen that
563 a downstream mirror did not see the add at all! It seems this needs to
564 be mentioned somewhere. The point here is that even if the downstream is
565 never missing the principal timeframe it may encounter a "delete" that
566 has no complimentary "add" anywhere.
568 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
570 * I suspect the treat of metadata is incorrect during read or something.
571 The bug that I am watching is that between 06:08 and 06:09 the 6h file
572 contained more than 6 hours worth of data. At 06:08 we merged into the
573 1d file. We need to take snapshots of the 6h file over the course of an
574 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
575 enough.
577 Much worse: watching the 1h file: right at the moment (at 06:35) it
578 covers 1218867584-1219120397 which is 70 hours.
580 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
581 08:19:44 2008, that is when I checked out last time, so it seems to be
582 aggregating and never truncating?
584 No, correct is: it is never truncating; but wrong is: it is aggregating.
585 It does receive a lot of events from time to time from a larger file.
586 Somehow a large file gets merged into the small one and because the
587 "meta/merged" attribute is missing, nobody is paying attention. I
588 believe that I can fix this by making sure that metadata are honoured
589 during read. DONE and test adjusted.
591 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
593 * grand renaming plan
595 remotebase => remoteroot to fit well with localroot DONE
596 local_path() => localroot seems to me should already work DONE
597 recentfile_basename => rfilename no need to stress it has no slash DONE
599 filenameroot??? Doesn't seem too bad to me today. Maybe something like
600 kern? It would anyway need a deprecation cycle because it is an
601 important constructor.
603 * I like the portability that Data::Serializer brings us but the price
604 is that some day we might find out that it is slowing us a bit. We'll
605 see.
607 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
609 * should we not enter the interval of the principal (or the interval of
610 the merging file?) in every aggregated/merged file?
612 * we should aim at a first release and give up on thinking about
613 sanitizing stuff and zloop. Let's just admit that a full traditional
614 rsync is the only available sanitizer ATM. Otherwise it's complicated
615 stuff: sanitizing on the origin server, sanitizing on the slaves,
616 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
617 get the basics out before this becomes a major cause for mess.
619 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
621 * On OSes not supporting symlinks we expect that RECENT.recent contains
622 the contents of the principal recentfile. Actually this is identical on
623 systems supporting symlinks. Simple, what follows from that is that we
624 need to keep the serializer in the metadata because we cannot read it
625 from the filename, doesn't it? Of course not. It's a chicken and egg
626 problem. This leaves us with the problem to actually parse the
627 serialized data to find out in which format it is. So who can do the 4
628 or 5 magics we wanted to support? File::LibMagic?
630 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
632 * remotebase and recentfile_basename are ugly names. Now that we need a
633 word for the shortest/principal/driving recentfile too we should do
634 something about it.
636 localroot is good. rfile is good. local_path() is bad, local_path($path)
637 is medium, filenameroot() is bad, remotebase is bad, recentfile is
638 already deprecated.
640 Up to now remotebase was the string that described the remote root
641 directory in rsync notation, like pause.perl.org::authors. And
642 recentfile_basename was "RECENT-1h.yaml".
644 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
646 * The test that was added in today's checkin is a good start for a test
647 of rmirror. We should have more methods in Recent.pm: verify,
648 addmissingfiles. We should verify the current tree, then rmirror it and
649 then verifytree the copy. We could then add some arbitrary file and let
650 it be discovered by addmissingfiles, then rmirror again and then
651 verifytree the copy again.
653 Then we could start stealing from csync2 sqlite database [no port to
654 OSX!] and fill a local DB. And methods to compare the database with the
655 recentfiles. Our strength is that in principle we could maintain state
656 with a single float. We have synced up to 1234567890.123456. If the Z
657 file does not add new files all we have to do is mirror the new ones and
658 delete the goners.
660 This makes it clear that we should extend current protocol and declare
661 that we cheat when we add files too late, just to help the other end
662 keeping track. Ah yes, that's what was meant when zloop was mentioned
663 earlier.
665 Maybe need to revisit File::Mirror to help me with this task.
667 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
669 * There must be an allow-me-to-truncate flag in every recentfile.
670 Without it one could construct a sequence of updates winning the locking
671 battle against the aggregator. Only if an aggregator has managed to
672 merge data over to the next level, truncating can be allowed. DONE with
673 accessor merged.
675 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
677 * We should probably guarantee that no duplicates enter the aggregator
678 array.
680 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
682 * To get merge operation faster would need a good benchmark test. What
683 02 spits out isn't reliable enough and is dominated by many other
684 things. Between
686 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
687 Author: Andreas J. Koenig <k@k75.(none)>
688 Date: Sat Aug 2 07:58:04 2008 +0200
690 and
692 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
693 Author: Andreas J. Koenig <k@k75.(none)>
694 Date: Sat Aug 2 11:40:29 2008 +0200
696 there was a lot of trying but the effect is hardly measurable with
697 current tests.
699 * overhead of connecting seems high. When setting
700 max_files_per_connection to 1 we see that.
702 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
704 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
706 25138.2215600014/3600 = 6.98283932222261
708 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
709 be a bug in the aggregator. FIXED
711 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
713 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
714 we have it, can we mirror it, mirror it!
716 I fear this needs a new class which might be called
717 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
718 and treat them as an entity. I realize that a single recentfile may be
719 sufficient for certain tasks and that it is handy for the low level
720 programmer but it is not nice to use. If there is a delete in the 1h
721 file then the 6h file still contains it. Seekers of the best information
722 need to combine at least some of the recentfiles most of the time.
724 There is the place for the Z loop!
726 But the combination is something to collect in a database, isn't it. Did
727 csync2 just harrumph?
729 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
731 * it just occurred to me that hosts in the same mirroring pool could
732 help out each other even without rewriting the recentfile. Just fetch
733 the stuff to mirror from several places, bingo. But that's something
734 that should rather live in a separate package or in rsync directly.
736 * cronjobs are unsuited because with ntp they would all come at the full
737 minute and disturb each other. Besides that I'd hate to have a backbone
738 with more than a few seconds latency.
740 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
742 * a second rsync server with access control for PAUSE. Port? 873 is the
743 standard port, let's take 8873.
745 * if there were a filesystem based on this, it would have a slow access
746 to inexistent files. It would probably provide wrong readdir (only based
747 on current content) or also a slow one (based on a recentfile written
748 after the call). But it would provide fast access to existing files. Or
749 one would deliberately allow slightly blurred answers based on some
750 sqlite reflection of the recentfiles.
752 * todo: write a variant of mirror() that combines two or more
753 recentfiles and treats them like one
755 * todo: signal handler to remove the tempfile
757 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
759 * now that we have the symlink I forgot how it should be used in
760 practice.
762 * the z loop: add missing files to Z file. Just append them (instead of
763 prepending). So one guy prepends something from the Y file from time to
764 time and another guy appends something rather frequently. Collecting
765 pond. When Y merges into Z, things get epoch and the collecting pond
766 gets smaller. What exactly are "missing files"?
768 take note of current epoch of the alpha file, let's call it the
769 recent-ts
771 find all files on disk
773 remove all files registered in the recentworld up to recent-ts
775 remove all files that have been deleted after recent-ts according to
776 recentworld
778 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
780 * rersyncrecent might be a cronjob with a (locked) state file which
781 contains things like after and maybe last z sync or such?
783 rrr-mirror might be an alternative name but how would we justify the
784 three Rs when there is no Re-Rsync-Recent?
786 With the --loop parameter it is an endless loop, without it is no loop.
787 At least this is simple.
789 * todo: new accssor z-interval specifies how often the Z file is updated
790 against the filesystem. We probably want no epoch stamp on these
791 entries. And we want to be able to filter the entries (e.g. no
792 by-modules and by-category tree)
794 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
796 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
797 into Z. We do not need the other files filled up with filesystem
798 contents though.
800 * need interface to query for a file in order to NOT call update on
801 PAUSE a second time within a short time.
803 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
805 * recommended update interval? Makes no sense, is different for
806 different users.
808 * Moosify
810 Local Variables:
811 mode: change-log
812 change-log-default-name: "Todo"
813 tab-width: 2
814 left-margin: 2
815 End: