has anyone here had to load balance PACS/DICOM tra...
# citrix-netscaler
m
has anyone here had to load balance PACS/DICOM traffic before? Looking for guidance on monitors, persistence settings, TCP Profile settings.
b
I have for Synapse and Sectra ones are in progress.
m
do you use a custom monitor?
I'm running into an issue where after a reboot the TCP monitor sees the port as up but users will get sent to a backend server that won't accept traffic and they are stuck there due to long session persistence timeouts recommended by the vendor
b
we do, the vendors provided info on the monitors
m
Did you do it with a custom script or what? And what are you using for persistence? Any specifics on the tcp profile?
b
Custom monitors. For Synapse it connects to a specific port. For Sectra it checks for a 200 response for a specific URL.
l
we used some version of this as a custom monitor : loadbalancer.org/blog/load-balancing-dicom-pacs-health-check you can use it to query actual aetitles for responsiveness. tcp profile: had a lot of throughput complaints when initially migrating from another type of LB to netscaler, until we tweaked some settings, dicom is a really chatty protocol, using so much chatter to send through actual data. started off with a default lan profile and adjusted: max burst limit -> 255 initial congestion win size -> 20 max packets per retransmission -> 50 slow start increment -> 25 This was early 2016, so pretty sure we had to enable SACK aswell back then, but that seems to be ok in todays standards :) Persistency here is just a 25 minute source ip one, that was the original setting on the old config as well
👍 1