[
https://issues.apache.org/jira/browse/IGNITE-20122?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Roman Puchkovskiy updated IGNITE-20122:
---------------------------------------
Description:
When a DROP INDEX is executed for an index that is STARTING or BACKFILLING, it
should be removed right away (skipping the AVAILABLE and subsequent states) and
start destruction. This should be done using a conditional schema update
(IGNITE-20115) to avoid a race with switching to the BACKFILING/AVAILABLE state.
If the conditional schema update fails (because the index has been switched to
the BACKFILLING/AVAILABLE state), we should retry at the new state. If it's
BACKFILLING, we should try to destroy the index again; if it's AVAILABLE, then
fallback to the usual procedure (IGNITE-20119).
was:
When a DROP INDEX is executed for a an index that is STARTING, it should be
removed right away (skipping the READY and STOPPING states) and start
destruction. This should be done using a conditional schema update
(IGNITE-20115) to avoid a race with switching to the READY state.
If the conditional schema update fails (because the index has been switched to
the READY state), we should fallback to the usual procedure (IGNITE-20119).
> When a REGISTERED/BACKFILLING index is dropped, it should be removed right
> away
> -------------------------------------------------------------------------------
>
> Key: IGNITE-20122
> URL: https://issues.apache.org/jira/browse/IGNITE-20122
> Project: Ignite
> Issue Type: Improvement
> Reporter: Roman Puchkovskiy
> Priority: Major
> Labels: ignite-3
> Fix For: 3.0.0-beta2
>
>
> When a DROP INDEX is executed for an index that is STARTING or BACKFILLING,
> it should be removed right away (skipping the AVAILABLE and subsequent
> states) and start destruction. This should be done using a conditional schema
> update (IGNITE-20115) to avoid a race with switching to the
> BACKFILING/AVAILABLE state.
> If the conditional schema update fails (because the index has been switched
> to the BACKFILLING/AVAILABLE state), we should retry at the new state. If
> it's BACKFILLING, we should try to destroy the index again; if it's
> AVAILABLE, then fallback to the usual procedure (IGNITE-20119).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
