Dear Heikki

This is great, your suggestions solved the problem!

> When AllowAuthorizeOnly is set, Radiator triggers an Access-Request that 
> has 'Service-Type = Authorize-Only' but no User-Password attribute. In 
> your case you could catch these requests with a specific Handler and 
> then run the 'authorizeSQL' AuthBy only within this new Handler.
> 
> When you know you can handle 'Service-Type = Authorize-Only' TACACS+ 
> derived access requests, you can enable FarmSize on the frontend.

With the first step Authorization-Only in the backend and a FarmSize of 8 in 
the frontend the TCP errors dropped from approx. 1000 to 100 per second. With a 
doubling of the FarmSize to 16 they decreased again but stayed on a level of 
approx. 50/s. 

We observed that the CPU load was still on 100% on all cores for 2-3 seconds. 
Thus we also doubled the count of virtual CPUs from 8 to 16 and with this step 
the errors are finally gone. The "tacacs server unreachable" logs on the 
clients (switches & routers) have also disappeared completely.

This is the config we added:

FRONTEND (before <Client ...> section:
FarmSize 16
DupCache shared
DupCacheFile /var/run/radius/rad_auth-tacacs-frontend-%0

BACKEND (before default <Handler> section:
<Handler Service-Type=Authorize-Only>
Identifier TacacsAuthorizeOnly
AuthByPolicy ContinueWhileAccept
AuthBy SQLauthorizeTAC
AuthBy InternalReply
RejectHasReason
AuthLog authlog-tacacs
</Handler>

Thank you and best regards, 
Tobias

-------------------------------------------------------
ETH Zürich
Tobias Schnurrenberger
ID INFRA Network Applications
Binzmühlestrasse 130
8092 Zürich

[email protected]
-------------------------------------------------------

Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
radiator mailing list
[email protected]
https://lists.open.com.au/mailman/listinfo/radiator

Reply via email to