ignite-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Pavel Kovalenko (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (IGNITE-9420) Move logical recovery phase outside of PME
Date Thu, 01 Nov 2018 09:38:00 GMT

    [ https://issues.apache.org/jira/browse/IGNITE-9420?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16671357#comment-16671357
] 

Pavel Kovalenko commented on IGNITE-9420:
-----------------------------------------

[~agoncharuk] [~DmitriyGovorukhin]
Changes are ready. Summary:
# Logical recovery is moved outside of PME as next stage of binary recovery implemented there
IGNITE-7196
# In logical recovery stage WAL resumes logging but allows to log only Page Delta and Page
Snapshot records.
# Before logical updates will be applied all locally available caches start in special "recovery
mode". In this mode cache group offheap manager can apply cache updates from WAL and create/restore
partitions. Other parts that require discovery (preloading, affinity fetching) are off.
# After logical updates were applied all cache groups restore their latest partition states
and WAL manager is deactivated.
# First PME (node join to active cluster or activation) finishes recovery for all caches and
groups that were started in the recovery phase enabling preloading, affinity fetching and
attaching start topology version.

> Move logical recovery phase outside of PME
> ------------------------------------------
>
>                 Key: IGNITE-9420
>                 URL: https://issues.apache.org/jira/browse/IGNITE-9420
>             Project: Ignite
>          Issue Type: Improvement
>          Components: cache
>    Affects Versions: 2.5
>            Reporter: Pavel Kovalenko
>            Assignee: Pavel Kovalenko
>            Priority: Major
>              Labels: cache
>             Fix For: 2.8
>
>
> Currently, we perform logical recovery in PME here org.apache.ignite.internal.processors.cache.persistence.GridCacheDatabaseSharedManager#restoreState
> We should move logical recovery before discovery manager will start.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message