cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Robert Coli (JIRA)" <>
Subject [jira] [Commented] (CASSANDRA-1053) Expose setting of phi in the FailureDetector
Date Thu, 18 Oct 2012 18:08:03 GMT


Robert Coli commented on CASSANDRA-1053:

Why are the values here clamped in the conf file and not in JMX? That seems like a peculiar
pattern, where the same config variable has two potential ranges for values. Is there some
design pattern or process I am not familiar with which recommends this? Because it seems to
have really trivial pathological and confusing cases.. for example, how would a document describe
the range?
> Expose setting of phi in the FailureDetector
> --------------------------------------------
>                 Key: CASSANDRA-1053
>                 URL:
>             Project: Cassandra
>          Issue Type: Improvement
>            Reporter: Brandon Williams
>            Assignee: Brandon Williams
>            Priority: Trivial
>             Fix For: 0.6.2
>         Attachments: 1053-trunk.txt, 1053.txt
> I've seen some users, always on cloud platforms, say that they have problems with hosts
flapping in the FD.  GC is not the cause.  I know of at least one production deployment where
they are already hacking phi to 10 to get around the flapping problems.
> This is a dangerous thing to expose, however, since giving meaning to the difference
of 8 and 10 here is hard to quantify to someone unfamiliar with the inner workings.  Perhaps
we can allow setting it in the config, but not specify a default and leave it commented out
with a big fat warning.

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see:

View raw message