Additional heartbeat logger
---------------------------
Key: SEQUOIA-920
URL: https://forge.continuent.org/jira/browse/SEQUOIA-920
Project: Sequoia
Type: Improvement
Components: Core
Versions: Sequoia 2.10.6, Sequoia 2.10.5, Sequoia 3.0 beta2, Sequoia
2.10.4, Sequoia 2.10.3, Sequoia 3.0 beta1, Sequoia 2.10.2, Sequoia 2.10.1,
Sequoia 2.10, Sequoia 2.9
Reporter: Marc Herbert
There is a strong desire for an additional "heartbeat" logger, exactly
like syslogd (see "-- MARK --" in syslogd man page). This would solve
(at least) two problems:
- when reading the logs, you find a 5 hours hole before controller
restart. When did the host crashed approximately?
- under heavy load mysterious time behaviours have been noticed,
including clock drifts (!) under linux.
Design sketch:
- has its own, new dedicated logger
- period is configurable through the debug level (e.g.:
DEBUG: every 15seconds, INFO: every 2minutes, etc.)
- lifecycle = controller lifecycle
- logs lines like this:
2007-01-12 19:21:36,247 --- BEAT --- controller has been up for HH:MM:SS
HH:MM:SS is _NOT_ computed using wall-clock, but by multiplying the
the number of MARKs by the sleep period instead => trivial detection
of clock drifts issues
java.util.Timer/TimerTask seem well-suited for the job (we have enough
bare threads to manage already).
Please provide comments/make suggestions. Especially:
- alternatives names more intuitive for native english speakers
- good default period values, preferably inspired by field experience
Marc +Jeff
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
https://forge.continuent.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see:
http://www.atlassian.com/software/jira
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
