Actions
Architecture #11906
closed
rsyslog handling of received log on rudder server is inefficient
Pull Request:
Effort required:
Name check:
Fix check:
Regression:
Description
On the Rudder server, each time rsyslog receives a message, it writes it in the database, doing a transaction for only one log message
This is highly inefficient, both for rsyslog and database, and may lead to massive slowdown of infra if rsyslog is used in tcp
Some pointer on how to do it there:
http://www.rsyslog.com/doc/v8-stable/concepts/queues.html
http://www.rsyslog.com/doc/v8-stable/rainerscript/queue_parameters.html
http://rsyslog-users.1305293.n2.nabble.com/Rsyslog-Disk-Queue-Flush-Issue-td7580422.html
- how can we force the message to get fast to the database, for it to be handled by Rudder and not lag there for 1 hours if there are few nodes ?
- memory usage of storing in memory logs ?
Updated by 
Updated by 
Updated by
Actions