[
https://issues.apache.org/jira/browse/UNOMI-1025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jonathan Sinovassin-Naïk updated UNOMI-1025:
--------------------------------------------
Fix Version/s: unomi-3.1.0
(was: unomi-3.0.2)
> Healthcheck: restore the provider timeout
> -----------------------------------------
>
> Key: UNOMI-1025
> URL: https://issues.apache.org/jira/browse/UNOMI-1025
> Project: Apache Unomi
> Issue Type: Bug
> Components: unomi(-core)
> Reporter: Jerome Blanchard
> Priority: Major
> Fix For: unomi-4.0.0, unomi-3.1.0
>
> Time Spent: 40m
> Remaining Estimate: 0h
>
> Links: relates to UNOMI-924
> UNOMI-924 (https://github.com/apache/unomi/pull/746) makes the healthcheck
> endpoint accept concurrent calls. The change also removes the per-provider
> timeout. HealthCheckService calls provider.execute() directly, and the
> "timeout" configuration key is no longer read.
> The HTTP client of the ElasticSearch and OpenSearch providers has no connect
> timeout and no socket timeout, because HttpUtils.initHttpClient uses an empty
> RequestConfig. If the search engine stops to answer but keeps the TCP
> connection open, these problems occur:
> * The refresh thread waits with no limit, and "busy" stays true.
> * Every other call gets the last cached result with no age limit. If that
> result was LIVE, /health/check returns HTTP 200 while the search engine does
> not answer.
> * Before the first refresh completes, every call waits on cacheLock with no
> limit.
> Two minor problems are in the same class:
> * Two threads can pass shouldRefreshCache() at the same time. The second
> thread then runs a second refresh after the first, because it does not check
> the cache age again inside cacheLock.
> * The providers list is an ArrayList that bind() and unbind() change while
> refreshCache() reads it.
> Proposed fix:
> # Run each provider call in an executor, and wait for the result with the
> configured "timeout" value. On timeout, cancel the call and return
> provider.timeout().
> # Set the connect and socket timeouts of the ElasticSearch and OpenSearch
> providers to the same "timeout" value, so that a blocked call ends.
> # Check the cache age again inside cacheLock.
> # Use a CopyOnWriteArrayList for the providers list.
> A unit test with a provider that never answers checks that check() returns
> within the timeout and marks that provider as timed out.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
