[
https://issues.apache.org/jira/browse/TIKA-2889?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16857814#comment-16857814
]
Thomas van Hesteren commented on TIKA-2889:
-------------------------------------------
Yes, something happened... the /tika request is the life-checker. As you can
see in my previous message, my script detects a wrong cUrl response from
parsing a file. This results in the life-check to be performed (/tika request).
And this indicated that Tika wasn't running anymore, which results in the
reboot for Tika server. Strange thing however is that in the logging you see
the 'This is Tika Server' response. Though, this check was not valid:
if (cUrlResponseTika.substr(0, 12) == "This is Tika") {
return true;
}
else {
return false;
}
So, how is it possible that the debug logs are showing a success response while
the response wasn't correct? Besides, that's something to look into regarding
the life-check. However, that the life-check is triggered at all is wrong
already...
As already stated above, the life-check is only initiated when I get a
curl_code != CURLE_OK from the cUrl request parsing a file.
What do you advise me to do in order to debug this issue further?
> Tika Server keeps crashing
> --------------------------
>
> Key: TIKA-2889
> URL: https://issues.apache.org/jira/browse/TIKA-2889
> Project: Tika
> Issue Type: Bug
> Components: server
> Affects Versions: 1.18, 1.19, 1.19.1, 1.21
> Environment: Both Ubuntu and Windows have the same bug/issue
> Reporter: Thomas van Hesteren
> Priority: Minor
> Attachments: log4j.xml, tika-2.log, tika-server-everything-2.log,
> tika-server-everything.log, tika-server-everything.log,
> tika-server-everything.log, tika.log, tika.log
>
>
> I have a document processor which sends documents to the Tika Server over
> cUrl. However, the server crashes multiple times (not document specific). The
> response I get from cUrl if it happens is as follows:
> Connection error: Couldn't connect to server
>
> The Tika server is started when the script starts executing. For now, I fixed
> the issue by making a watcher which restarts the tika server when it crashes.
> It then processes a few other documents and crashes again (after a few
> minutes, let's say 5 minutes tops).
>
> Is there any possibility to catch the exception (if it throws any?)
>
> A log which shows the crash of the server:
> 04-06-2019 15:49:25|Processing a file of: 52.3kB
> 04-06-2019 15:49:24|Processing a file of: 255.5kB
> 04-06-2019 15:49:24|Processing a file of: 241.6kB
> 04-06-2019 15:49:23|Processing a file of: 37.7kB
> 04-06-2019 15:49:22|Processing a file of: 1.27MB
> 04-06-2019 15:49:21|Processing a file of: 55.8kB
> 04-06-2019 15:49:17|Processing a file of: 114.5kB
> 04-06-2019 15:49:08|Server is not running. Restarting Server. Connection
> error: Couldn't connect to server
> 04-06-2019 15:49:03|Processing a file of: 41.0kB
> 04-06-2019 15:49:00|Processing a file of: 38.0kB
> 04-06-2019 15:48:59|ProcesPsing a file of: 37.1kB
> 04-06-2019 15:48:59|Processing a file of: 60.2kB
> 04-06-2019 15:48:59|Processing a file of: 280.7kB
> 04-06-2019 15:48:59|Processing a file of: 3.30MB
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
