Thank you AUSNOG community for all the replies, I appreciate all the input and
suggestions.
I've narrowed this down to a carrier issue as when we preference away the
problem goes away.
I will follow up and sort.
Thanks again.
Darren.
From: AusNOG On Behalf Of Darren Moss
Sent: Wednesday, 19 January 2022 4:38 PM
To: AusNOG Mailing List
Subject: [AusNOG] Challenge with SQL Replication over multiple carriers
Hello Noggers,
I have a handful of customers with SQL replication issues over 2 separate
transit links between Sydney and Melbourne.
Looking to see if anyone here has experienced something similar or could
potentially shed some light, perhaps more on my carrier configuration and
traffic preferencing.
We have 2x 10GBps interfaces in 2 separate datacentres for transit across 2
carriers. This link carries SQL replication.
It appears when we are using 1 carrier link, everything seems fine and traffic
replicates normally.
However, when we balance between 2 carriers, into the same networks, we start
seeing issues.
Connections are working, I can telnet, ICMP, etc, however when we start sending
large packets we see semaphore timeout errors.
To me it points to TCPOffload, however that's all disabled and I can see no
incorrect packets, etc at both sides.
I have even tried routing the SQL traffic over a different network within the
same BGP sessions, which works for a while, then eventually starts barfing with
semaphore timeout issues.
My question is with traffic preferencing, is there anything special I need to
do when using multiple carriers at the same time for the same networks ?
Happy to hear input and suggestions, off list would be better. I am happy to
publish what steps were taken.
Many thanks
Darren.
___
AusNOG mailing list
AusNOG@lists.ausnog.net
http://lists.ausnog.net/mailman/listinfo/ausnog