> On May 14, 2014, 10:55 p.m., Kevin Sweeney wrote: > > src/main/python/apache/aurora/client/config.py, line 161 > > <https://reviews.apache.org/r/21297/diff/4/?file=581994#file581994line161> > > > > How about adding a couple-second delay to account for executor bootup > > time. Ideally we'd be able to know when the executor started its stopwatch > > but in absence of that information we can fuzz it with a few second sleep.
I am afraid that would be hard to explain and is not quite deterministic. There is always a delay in status propagation that can wary from a few seconds to infinity. It's hard to say if 2 seconds is going to cover the majority of all cases. I think we should raise awareness here rather than provide any guarantees. How about making it "greater than" instead of "at least" requirement? That would give us at least 1 second slack on the watch side and stimulate users to think harder about their values. Here is the updated test output: ...CRITICAL:root: You have specified an insufficiently short watch period (10 seconds) in your update configuration. Your update will always succeed. In order for health notifications to be delivered, UpdateConfig.watch_secs must be greater than 15 seconds to account for an initial health check interval (15 seconds) plus 0 consecutive failures at a check interval of 10 seconds. .CRITICAL:root: You have specified an insufficiently short watch period (25 seconds) in your update configuration. Your update will always succeed. In order for health notifications to be delivered, UpdateConfig.watch_secs must be greater than 25 seconds to account for an initial health check interval (15 seconds) plus 1 consecutive failures at a check interval of 10 seconds. - Maxim ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/21297/#review43048 ----------------------------------------------------------- On May 14, 2014, 10:35 p.m., Maxim Khutornenko wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/21297/ > ----------------------------------------------------------- > > (Updated May 14, 2014, 10:35 p.m.) > > > Review request for Aurora, Kevin Sweeney and Brian Wickman. > > > Bugs: AURORA-404 > https://issues.apache.org/jira/browse/AURORA-404 > > > Repository: aurora > > > Description > ------- > > Adding checks for watch_secs, initial_interval_secs and restart_threshold. > > > Diffs > ----- > > src/main/python/apache/aurora/client/config.py > 350d84c97fac9f1dd834bf509a160faaacb26386 > src/main/python/apache/aurora/config/__init__.py > 74e6b21ef905221cb3beb974a540325937480f2e > src/main/python/apache/aurora/config/schema/base.py > 61a6680f8b1a463055ccd6d318cc540aca166684 > src/test/python/apache/aurora/client/cli/util.py > e17f256fc5d17f251e8f2cb4d94233d2d6c09897 > src/test/python/apache/aurora/client/commands/util.py > c28096fe89c9af6d5518bc0f8f871e886ca12e7f > src/test/python/apache/aurora/client/test_config.py > 8ef08685c317c3f9dae799dfb6bdced7077a8778 > > Diff: https://reviews.apache.org/r/21297/diff/ > > > Testing > ------- > > ./pants src/test/python:all > > > Thanks, > > Maxim Khutornenko > >
