geode-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mario Ivanac (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (GEODE-7039) Server recovery severely degrades client read traffic (no SingleHop no TX) on redundant partitioned persistent regions
Date Fri, 16 Aug 2019 06:52:00 GMT

     [ https://issues.apache.org/jira/browse/GEODE-7039?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Mario Ivanac updated GEODE-7039:
--------------------------------
    Description: Client not using single hop nor transactions is experiencing severe throttling
from the cluster when getting data from a partitioned persistent region while server hosting
one of the redundant buckets is recovering (in the process of image recovery). Get operation
that have not landed on a server hosting the bucket will be proxied to other members that
do have the bucket in a random fashion. This random picking has the nasty consequence that
chosen server might be the one recovering now and the bucket is not yet ready (BucketNotFoundException),
which means local server will handle ForceReattemptException by sleeping 100ms before another
(random) attempt. This sleeping is devasteting for throughput observed by the client.  (was:
Client not using single hop nor transactions is experiencing severe throttling from the cluster
when getting data from a partitioned region while server hosting one of the redundant buckets
is recovering (in the process of image recovery). Get operation that have not landed on a
server hosting the bucket will be proxied to other members that do have the bucket in a random
fashion. This random picking has the nasty consequence that chosen server might be the one
recovering now and the bucket is not yet ready (BucketNotFoundException), which means local
server will handle ForceReattemptException by sleeping 100ms before another (random) attempt.
This sleeping is devasteting for throughput observed by the client.)

> Server recovery severely degrades client read traffic (no SingleHop no TX) on redundant
partitioned persistent regions
> ----------------------------------------------------------------------------------------------------------------------
>
>                 Key: GEODE-7039
>                 URL: https://issues.apache.org/jira/browse/GEODE-7039
>             Project: Geode
>          Issue Type: Improvement
>          Components: client/server
>            Reporter: Mario Ivanac
>            Assignee: Mario Ivanac
>            Priority: Major
>
> Client not using single hop nor transactions is experiencing severe throttling from the
cluster when getting data from a partitioned persistent region while server hosting one of
the redundant buckets is recovering (in the process of image recovery). Get operation that
have not landed on a server hosting the bucket will be proxied to other members that do have
the bucket in a random fashion. This random picking has the nasty consequence that chosen
server might be the one recovering now and the bucket is not yet ready (BucketNotFoundException),
which means local server will handle ForceReattemptException by sleeping 100ms before another
(random) attempt. This sleeping is devasteting for throughput observed by the client.



--
This message was sent by Atlassian JIRA
(v7.6.14#76016)

Mime
View raw message