Receive "Timeout waiting for connection from pool" when attempting to connect to Atlassian Marketplace

Question

Hello,

When I try and connect to the Atlassian Marketplace from my production JIRA server, the connection fails and this is printed in the Catalina.out log:

2017-02-22 17:01:38,745 http-bio-8080-exec-350 WARN username 1021x360110x1 18ng3xd 10.106.149.41 /plugins/servlet/upm/marketplace [atlassian.upm.pac.PacClientImpl] Error when querying application info from MPAC: com.atlassian.marketplace.client.MpacException: org.apache.http.conn.ConnectionPoolTimeoutException: Timeout waiting for connection from pool

Ping to marketplace.atlassian.com works successfully:

[log]$ ping marketplace.atlassian.com
PING plugins.atlassian.com (131.103.28.7) 56(84) bytes of data.
64 bytes from pillar03.atlassian.com (131.103.28.7): icmp_seq=1 ttl=47 time=228 ms
64 bytes from pillar03.atlassian.com (131.103.28.7): icmp_seq=2 ttl=47 time=223 ms
64 bytes from pillar03.atlassian.com (131.103.28.7): icmp_seq=3 ttl=47 time=223 ms
^C
--- plugins.atlassian.com ping statistics ---
3 packets transmitted, 3 received, 0% packet loss, time 2869ms
rtt min/avg/max/mdev = 223.251/225.190/228.935/2.648 ms

I can connect successfully from my JIRA DEV box (running either JIRA 6.1.7 or JIRA 7.2.4). My production box is running JIRA 6.1.7. Both PRD and DEV run on a RH Linux VM.

I did some searching online, which indicated the error is a configuration error, but I'm not sure where I can configure this.

Appreciate any help in advance.

Daniel S1 · Answer

Thanks Mark.

I did a diff between the two JIRA install directories (DEV vs PRD) and they are pretty much the same except for the files under the work directory. They are so different between the DEV and PRD instances that I assume these files are added/updated at runtime. Is my understanding correct?

mlassau_atlassian · Answer

I can connect successfully from my JIRA DEV box (running either JIRA 6.1.7 or JIRA 7.2.4)

Hmmm ... I misread this at first as you can telnet from your dev box, but you mean that your test instance works fine.
That means you can probably rule out performance problems in Marketplace, or in your internet connection.

About the only thing you can do that might help out is try rebooting JIRA. (if there is a leak, then a reboot will fix it for some amount of time before you need to reboot again).

There is unlikely to be any configuration you can alter without recompiling source code. It is most likely a bug (eg hard-coded misconfiguration) and so upgrading to v7.x might already have a fix ... or at least you can then report the bug and upgrade to the fixed version some time later.

mlassau_atlassian · Answer

UPM is obviously using a HTTP Connection Pool (eg Apache HttpClient). The problem is not that JIRA cannot reach marketplace, but that the pool is exhausted.

That is, you either have too many concurrent threads trying to call marketplace, or each call is so slow that the connections are returned to the pool to slow, or the pool is too small (misconfigured as you suggest), or there is a connection pool leak, or ...

I want to tell you to raise a support ticket, but I see you are running v6.1.x - unsupported.
So instead I will suggest you try upgrading to a supported version of JIRA.