--On Friday, February 5, 2021 9:17 AM +0100 Frédéric Goudal <[email protected]> wrote:
Hello, I have a strange problem of synchronization, since yesterday my slave servers have a small difference of 2 or 3 minutes with the master. I mean the context CSN of the slave is 3 mn earlier than contextCSN of the master. My first question is the following : how the timestamp of the contextCSN of the slave is computed ? Is it a copy of the tilmestamp from the master ? Or is it based on the localtime of the slaves ? My second question is that maybe such a small difference is « normal » (as the contextCSN of slave is progressing that means that replication is not fully stopped just a bit delayed) and I should only worry for a greater difference of time than one minute.
CSN stands for change sequence number. It is composed of multiple parts. One part is the timestamp at which point the change was written on the provider, another is the provider that received the change, etc (particularly important for multi provider replication).
The CSN stored on the consumer is an indication of the last change sequence it processed from the provider.
With OpenLDAP it is critical that time be tightly synchronized between all nodes. If you have clock skew, you will end up with broken replication.
A 2-3 minute lag in replication would not generally be considered normal unless your systems had extremely heavy write load.
Regards, Quanah -- Quanah Gibson-Mount Product Architect Symas Corporation Packaged, certified, and supported LDAP solutions powered by OpenLDAP: <http://www.symas.com>
