[
https://issues.apache.org/jira/browse/OAK-9721?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17513130#comment-17513130
]
Mohit Kataria commented on OAK-9721:
------------------------------------
The importindex method consists of following steps:
# switchLanes : This step changes names of indexing lanes so that current
indexing don't alter this index. Assuming current indexing lane is "async ",
property changes from "async": "async" to "async":"temp-async" and
"async-previous":"async".
# importIndexData: In this step, index data is imported and indexes are marked
for disabling if index has supersedes property. At this stage we have a new
index committed with switchedlanes.
# bringIndexUpToDate: With this step we stop current indexing lane, find diff
between latest checkpoint and checkpoint we are using to import index.
# releaseCheckpoint: Release the old checkpoint we are using for importing.
As these states are persisted ones, we can't come back to exactly same
condition of nodestore in case the process halts/exits at any state. To
mitigate this to some extent, in case of failure, we revert back lane names
(there will be a window for which data is not indexed). But as we are saving
import state, we can figure out whether we should run import process again or
not.
> Add state information while indeximporting
> ------------------------------------------
>
> Key: OAK-9721
> URL: https://issues.apache.org/jira/browse/OAK-9721
> Project: Jackrabbit Oak
> Issue Type: Improvement
> Components: indexing
> Reporter: Mohit Kataria
> Assignee: Mohit Kataria
> Priority: Major
>
> While importing index (IndexImporter), there are majorly 4 steps. In case of
> failure the whole index becomes unusable.
> So to handle this, we are introducing a property to manage state of
> indeximport and add retry logic on each step.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
