karaf-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jean-Baptiste Onofré (JIRA) <j...@apache.org>
Subject [jira] [Commented] (KARAF-4874) Distributed log service no order of messages, no serialization
Date Wed, 10 May 2017 14:28:04 GMT

    [ https://issues.apache.org/jira/browse/KARAF-4874?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16004765#comment-16004765
] 

Jean-Baptiste Onofré commented on KARAF-4874:
---------------------------------------------

We define the following configuration for the log map:

{code}
    <map name="org.apache.karaf.cellar.log">
        <time-to-live-seconds>0</time-to-live-seconds>
        <max-idle-seconds>0</max-idle-seconds>
        <eviction-policy>LRU</eviction-policy>
        <max-size policy="PER_PARTITION">5000</max-size>
        <eviction-percentage>25</eviction-percentage>
        <backup-count>0</backup-count>
    </map>
{code}

Are you proposing to define backup count to 1 and increase the eviction percentage ?

Regarding the ordering, I guess you mean to order the ClusterLogKey (based on the timestamp
for instance) ?

> Distributed log service no order of messages, no serialization
> --------------------------------------------------------------
>
>                 Key: KARAF-4874
>                 URL: https://issues.apache.org/jira/browse/KARAF-4874
>             Project: Karaf
>          Issue Type: Bug
>          Components: cellar-core
>    Affects Versions: cellar-4.0.2
>            Reporter: Branislav Kalas
>            Assignee: Jean-Baptiste Onofré
>
> Cluster log writes records to hazelcast map with ClusterLogRecord instances as keys.
> cluster:log-display then just iterates through keyset of this map and outputs records
to console.
> Result is log messages with no ordering (no relation to order of putting records to map)
which i think is useless.
> When map reaches its max allowed size, there is LRU eviction strategy defined, which
is also incorrect (cause if you use cluster:display you use records withoud any order) and
record is just deleted, there is no persistence.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Mime
View raw message