Commit Graph

56 Commits

Author SHA1 Message Date
JoelKatz
51a85504a8 Startup and shutdown improvements. Performance improvements. 2013-03-03 02:09:19 -08:00
JoelKatz
167d13cf40 I think this is the underlying issue. In some cases where the tx return value wasn't
full success, we applied to our open ledger but didn't relay. This caused disputes
resolved only by dispute relaying.
2013-02-24 16:48:54 -08:00
JoelKatz
1e2aba6b38 Ledger acquire timing changes. Better logging of acquire stalls. 2013-02-21 02:47:19 -08:00
JoelKatz
52c378411d If we start seeing acquire timeouts, don't start new acquires.
The link could be overloaded.
2013-02-20 13:30:57 -08:00
JoelKatz
4d6bf6fd43 Logging reductions. 2013-02-19 15:41:03 -08:00
JoelKatz
a941a575ec Make sure we always have a "last validated ledger". 2013-02-17 16:49:37 -08:00
JoelKatz
a26a302476 Accept old ledgers acquired over the network at a lower priority. 2013-02-10 16:40:52 -08:00
JoelKatz
8fef4d9b59 Remove some extra logging. 2013-02-10 13:54:10 -08:00
JoelKatz
b6f9811001 Fix the bug that crashed ec3. 2013-02-09 20:07:36 -08:00
JoelKatz
4abdb34c20 Trivial cleanups. 2013-02-09 16:17:12 -08:00
JoelKatz
00fd3d5a49 Demote some logging. 2013-02-08 16:33:35 -08:00
JoelKatz
9781c10736 Fix some cases where ledger flags are incorrectly set. 2013-02-01 16:11:32 -08:00
JoelKatz
9a7ccf777e Fix a bug that can stall the acquire engine. 2013-01-28 01:13:39 -08:00
JoelKatz
e8ac00e1a3 Allow tunable node sizes to adjust cache sizes, sweep timing, fetch
timing, and so on. Node size defaults to "tiny", which converves
memory and bandwidth.
2013-01-27 13:55:59 -08:00
Arthur Britto
51bbe5ffc8 Add path test for issue #23. 2013-01-19 13:19:17 -08:00
JoelKatz
fb3e2e8af4 Fix a crash bug Arthur reported. 2013-01-15 17:09:57 -08:00
JoelKatz
9dea6df56b Tuning. 2013-01-14 18:16:47 -08:00
JoelKatz
55d1af746c Share the 'shouldAcquire' function. recvGetLedger will needed it. 2013-01-12 08:42:24 -08:00
JoelKatz
1b8f20eaa7 Some tuning. 2013-01-12 04:44:41 -08:00
JoelKatz
35721a73e2 Don't let the acquire engine stall. 2013-01-11 17:18:28 -08:00
JoelKatz
955c5c3a9c Bug fixes and performance improvements. 2013-01-11 16:40:37 -08:00
JoelKatz
e2bac0e7cd Optimize the startup code so we don't have long periods of slowness on startup. 2013-01-10 08:41:32 -08:00
JoelKatz
3bdba555eb Rate limit. 2013-01-10 07:04:35 -08:00
JoelKatz
cf284897bb Fix a case where we might try too quickly to re-acquire a ledger. 2013-01-09 15:55:41 -08:00
JoelKatz
38af346881 Fix some calls that bypass the cache. 2013-01-09 10:49:08 -08:00
JoelKatz
d2d84d2af7 More conservative check. 2013-01-09 09:53:39 -08:00
JoelKatz
9a2e2d78c6 Extra debug 2013-01-09 09:21:48 -08:00
JoelKatz
8e5ce2dd4c More debug. 2013-01-08 14:10:49 -08:00
JoelKatz
2a657a33c9 Make sure we leave need network ledger mode. 2013-01-08 00:41:07 -08:00
JoelKatz
69ac139485 Reduce log spew. 2013-01-08 00:36:13 -08:00
JoelKatz
7a2f098ac4 Reduce some chatty logging. 2013-01-07 23:47:21 -08:00
JoelKatz
ef2f653473 Set the accept flag for acquires triggered by the publish logic. 2013-01-07 09:01:31 -08:00
JoelKatz
2b479a66c1 Don't immediately retry a failed acquire. 2013-01-07 08:57:41 -08:00
JoelKatz
f493ea6478 Be more aggressive about avoiding publishing ledger holes. Make the logic simpler and more sensible. 2013-01-07 01:33:14 -08:00
JoelKatz
cf77fd88f9 Mark a FIXME. 2013-01-06 17:32:10 -08:00
JoelKatz
d998feb13c Don't start a publish thread if there's nothing to publish. 2013-01-04 15:23:03 -08:00
JoelKatz
d57b5a9797 Track failed acquires. Fix a case where an acquire both succeeds and fails. 2013-01-04 15:21:31 -08:00
JoelKatz
e683f92844 Fix one case where we stop acquiring historical ledgers after a failure. 2013-01-04 12:13:58 -08:00
JoelKatz
8ace87066a Fix a huge delay with --net 2013-01-04 01:10:13 -08:00
JoelKatz
f0c73a76ce Fix some local acquire logic 2013-01-04 00:40:03 -08:00
JoelKatz
06c54c425c Fix not leaving the "need network ledger" state correctly. 2013-01-04 00:08:27 -08:00
JoelKatz
b7fe1424fb Count a ledger as present (for purposes of allowing clients that trust us to
query it) only when it has been validated.
2013-01-03 21:04:11 -08:00
JoelKatz
1127ae560e Defer publishing a ledger to clients until it accumulates sufficient
validations. We now have an open ledger, a last closed ledger, and a last
validated ledger. Normally, a ledger will be validated right after it
closes, but in edge cases, we might see a ledger close that then doesn't get
validated. This makes the code do the right thing.
2013-01-02 12:04:16 -08:00
JoelKatz
b71d0a93f9 Bugfix. 2013-01-01 18:26:17 -08:00
JoelKatz
c2d654efee More of the new publication logic. 2013-01-01 18:21:19 -08:00
JoelKatz
1e24721a98 First part of the fix to stop us from publishing ledgers that don't get validated. 2013-01-01 16:11:30 -08:00
JoelKatz
8b8a538a5c On a ledger jump, properly invalidate prior ledgers and invalidate a backfill in progress. 2012-12-31 00:25:52 -08:00
JoelKatz
fa1db60011 Correctly check for acquire doneness. 2012-12-25 21:07:53 -08:00
JoelKatz
be2e55d49c Fix some cases where the acquire engine can stall. 2012-12-25 21:05:12 -08:00
JoelKatz
7c13c57638 Make sure we clean up on a failed acquire. 2012-12-24 20:48:54 -08:00