NetScaler Load Balancing recommendation

[Marek_Aksamit]

Hello,

Well, another post about load balancing. Yes, there are bunch of posts out there, but to be honest nothing is specific, and it looks like Hyland does not have any recommendation. I already went through the "Load balancing" white paper and still looking for answers. 

What setting are you using for load balancing and OnBase (with SMART on FHIR).

- we are on 23.1.7 (SoF)

- Web/app servers on the same box - Web server is configured for SOAP communication

-Persistance  - SourceIp (time out 20 min)

We've been experiencing huge IIS connection spikes to our web/app servers once in a while, and we can't really pinpoint the issue.  Anyone willing to share their LB setting specific to OnBase?

Thanks

Marek

[Joe_Bunch]

We are very similar except timeout settings.

Web/App on the same box, pointed to locally talk

Persistence IP (Source Address)

Timeout 84600 seconds.

So to make sure we are talking the same for talking path:

Epic -> LB -> Server #1 Web -> Server #1 App (talks only to same server in our environment)

** also minus all the idp and api stuff...

[David_Juhlin]

We have our Web and App servers segregated:

- Web servers have Persistence IP; timeout is at least 31 minutes (so the LB timeout is longer than the web site timeout)

- App servers have Cookie-based persistence, timeout is 0 (never)

I currently have idp/ids/api on our App servers, but plan to move them to their own (most likely maintaining the cookie persistence). 

We are in the middle of our SoF migration (V21).  We have it in test but not prod and are working on upgrading to the 'fixed' build to address ALL EPIC USERS issue.  SoF seems to introduce a lot more traffic (lots of small packets) directly to the app server, so that might lead to IIS spikes.  (Could the underlying issue be more on the server resource side than the Load Balancer side?) 

[Joe_Bunch]

@David Juhlin Do you see benefits from that setup? I always wondered why you would go back out to another load balancer for APP. I can only assume that it allows a smaller Web footprint and maybe allows all the CPU/Memory hogging APPs to be isolated on more servers...? So 2 web servers pointed to 4 app servers.

[David_Juhlin]

@Joe Bunch ,

Yes.  We have a LOT of activity against our four app servers and not so much from our two web servers.  That said, during IPUP upgrades or for troubleshooting, I can/will point to a specific app server to better control which servers are running different builds of the code, etc.