--On Thursday, March 02, 2017 1:08 PM +0200 Nick Milas nick@eurobjects.com wrote:
Since "only some systems support the customization of these values; the keepalive parameter is ignored otherwise", how can I make sure that this parameter has been configured successfully and is enabled on my system (CentOS 7)? I enabled "loglevel config stats sync"but I don't see any mention of configured syncrepl parameters during OpenLDAP start.
Hi Nick,
The keepalive parameter was added after the initial 2.4 release, and after the current sections on replication in the admin guide were written. It was created because I kept encountering issues at customer sites where load balancers & packet shapers were closing the syncrepl connection.
I use the following: keepalive=240:10:30
The 240 because most of the hardware causing problems used a 5 minute timeout, so 240 was nicely under that.
If setting this resolves your problem, then you have something in your network monitoring and severing connections.
Regards, Quanah
--
Quanah Gibson-Mount Product Architect Symas Corporation Packaged, certified, and supported LDAP solutions powered by OpenLDAP: http://www.symas.com
On 2/3/2017 5:59 μμ, Quanah Gibson-Mount wrote:
If setting this resolves your problem, then you have something in your network monitoring and severing connections.
I used (as I have noted):
keepalive=20:100:2
on one consumer and:
keepalive=120:10:30
on the other (which is closer to your suggestion).
Both are working fine until now; no more disconnects.
I am still puzzled how this behavior started. We are not using any load balancers and/or packet shapers. As I have described, the setup is exactly the same as it was before migrating to CentOS 7 and v2.4.44 (from CentOS 5 and v2.4.39), where we did not face such issues. It may be related to the VM hosting environment, as both consumer boxes are VMs under KVM.
Anyway, things seem to be OK with the keepalive settings.
Thanks again, Nick
openldap-technical@openldap.org
-
Nick Milas -
Quanah Gibson-Mount