[ckan-dev] SSL SYSCALL error: EOF detected

Wed May 23 07:32:30 UTC 2018

I have opted for a conservative solution, modifying the next
/etc/sysctl.conf parameters

net.ipv4.tcp_keepalive_time=60
net.ipv4.tcp_keepalive_probes=3
net.ipv4.tcp_keepalive_intvl=10

and the postgresql.conf parameters sugested by Karen

tcp_keepalives_idle = 60
tcp_keepalives_interval = 60
tcp_keepalives_count = 10

For now it seems that the system is working correctly and the error has not
occurred again since I modified these parameters. Today, users are expected
to try again. I will cross my fingers.

Thanks you all.
Rafa.

2018-05-21 22:26 GMT+02:00 Karen Turner <karen at catalyst.net.nz>:

> Hi
>
> I had a similar problem. It was the firewall dropping the connection
> between the webserver and the database server. Because of the way that the
> connection was dropped the database server didn’t get the network
> disconnect.
> In my case we were using postgresql and we were able to set the keep alive
> in the postgresql config.
>
> I don’t know SQL alchemy itself well enough to know if there is a
> parameter that should be passed to make it retry a session if it goes away.
> Or whether this is something that would be done at the CKAN level.
>
> Cheers
>   Karen
>
> ----
> Karen Turner
> Catalyst IT
> w: http://catalyst.net.nz
> p: +64 4  803 2375
>
>
>
>
> On 22/05/2018, at 12:25 AM, Ross Jones <ross at mailbolt.com> wrote:
>
> Hi,
>
> The last time I had this problem, it was because the connection through
> the DMZ was being killed after a short period of inactivity.  As getting
> the DMZ connection changed was apparently not feasible, I ended up using
> pgbouncer ( https://pgbouncer.github.io/ ) to
> <https://pgbouncer.github.io/> handle the reconnections for me.  It
> worked pretty well.
>
> For Redis, I'd recommend looking at your redis.conf to ensure there is a
> timeout on connections, otherwise redis will keep those disconnected
> client's sockets around forever, or at least until you run out of
> file-descriptors.  This happened to my project this morning with some
> connections from 70 days ago still recorded in redis even though the
> clients had been rebooted many times.
>
> Solr I'm not so sure about, as far as I know a connection isn't kept
> permanently open, but I could be wrong - hopefully somebody else will know.
>
> Regards
>
> Ross
>
>
> ckan-dev mailing list
> ckan-dev at lists.okfn.org
> https://lists.okfn.org/mailman/listinfo/ckan-dev
> Unsubscribe: https://lists.okfn.org/mailman/options/ckan-dev
>
>
>
> _______________________________________________
> ckan-dev mailing list
> ckan-dev at lists.okfn.org
> https://lists.okfn.org/mailman/listinfo/ckan-dev
> Unsubscribe: https://lists.okfn.org/mailman/options/ckan-dev
>
>

-- 
--------------------------------------------------
Rafael Salas Robledo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.okfn.org/pipermail/ckan-dev/attachments/20180523/60bb13cd/attachment-0002.html>