[
https://issues.apache.org/jira/browse/IGNITE-20646?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexander Lapin updated IGNITE-20646:
-------------------------------------
Description:
h3. Motivation
[Vladislav
Pyatkov|https://issues.apache.org/jira/secure/ViewProfile.jspa?name=vpyatkov]
has strong point that current CLOCK_SKEW aware implementation of
await/getPrimaryReplica should be removed, meaning that we should compare
expirationTimestamp with propagated one without error(CLOCK_SKEW). Same as for
waiting. Current CLOC_SKEW aware approach is specified in
org.apache.ignite.internal.placementdriver.PlacementDriver javadocs and
guarantees that if primary replica is visible from await on node A it'll be
visible thought get on any other node.
h3. Definition of Done
We should prepare full set of pros and cons for each option. Whether they are
correct? Which one is better from the performance point of view? Which one is
simpler?
was:
h3. Motivation
[Vladislav
Pyatkov|https://issues.apache.org/jira/secure/ViewProfile.jspa?name=vpyatkov]
has strong point that current CLOCK_SKEW aware implementation of
await/getPrimaryReplica should be removed, meaning that we should compare
expirationTimestamp with propagated one without error(CLOCK_SKEW). Same as for
waiting.
h3. Definition of Done
We should prepare full set of pros and cons for each option. Whether they are
correct? Which one is better from the performance point of view? Which one is
simpler?
> Consider using CLOCK_SKEW unaware await/getPrimaryReplica()
> -----------------------------------------------------------
>
> Key: IGNITE-20646
> URL: https://issues.apache.org/jira/browse/IGNITE-20646
> Project: Ignite
> Issue Type: Improvement
> Reporter: Alexander Lapin
> Priority: Major
>
> h3. Motivation
> [Vladislav
> Pyatkov|https://issues.apache.org/jira/secure/ViewProfile.jspa?name=vpyatkov]
> has strong point that current CLOCK_SKEW aware implementation of
> await/getPrimaryReplica should be removed, meaning that we should compare
> expirationTimestamp with propagated one without error(CLOCK_SKEW). Same as
> for waiting. Current CLOC_SKEW aware approach is specified in
> org.apache.ignite.internal.placementdriver.PlacementDriver javadocs and
> guarantees that if primary replica is visible from await on node A it'll be
> visible thought get on any other node.
> h3. Definition of Done
> We should prepare full set of pros and cons for each option. Whether they are
> correct? Which one is better from the performance point of view? Which one is
> simpler?
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
