[
https://issues.apache.org/jira/browse/IGNITE-13858?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17306501#comment-17306501
]
Stanislav Lukyanov commented on IGNITE-13858:
---------------------------------------------
Talked about this with [~agoncharuk].
It seems that there is a relatively cheap way of improving this.
First, we can throw away the value from the entry stored on-heap. We only need
the key for the near cache algorithm to work.
Second, we can have a background thread iterating over the entries and removing
the ones that don't have any alive readers (near cache clients).
Third, we can have the same background thread perform no-op updates (e.g. no-op
EntryProcessor invocations) which will automatically remove unneeded entries
from the heap using the existing protocol.
> Enabling near cache on a client node may lead to blocking eviction on server
> nodes
> ----------------------------------------------------------------------------------
>
> Key: IGNITE-13858
> URL: https://issues.apache.org/jira/browse/IGNITE-13858
> Project: Ignite
> Issue Type: Bug
> Affects Versions: 2.8, 2.9, 2.9.1
> Reporter: Vyacheslav Koptilin
> Priority: Major
>
> When Native Persistence is off, creating near-cache on client node enforces
> (to be more accurate, near-cache entries) to maintain a list client node ids
> (aka readers) per-entry basis.
> Unfortunately, this list is not cleaned even though the eviction policy is
> configured on the near cache. Moreover, such entries cannot be evicted on the
> server-side, with corresponding pages, and this may lead to hanging cache
> operations.
> Originally discussed at user-list:
> http://apache-ignite-users.70518.x6.nabble.com/Apache-Ignite-Clientside-NearCache-and-Serverside-Eviction-blocking-cluster-tt34645.html
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
