A Cisco Identity Services Engine (ISE) node intermittently stops responding to RADIUS authentication requests, causing network devices to mark it as "RADIUS DEAD". The affected node receives RADIUS requests (confirmed via firewall logs and packet captures) but fails to respond. The following symptoms are observed:
RADIUS retransmission messages on network devices
"Radius Context Allocation Failed" critical alarms appearing in ISE
Log messages indicating "Hit context count limit. Dropping request!"
ISE GUI becomes inaccessible during the issue despite all services showing as running
Node restarts temporarily resolve the issue for approximately 30 minutes before it recurs
Specific error messages observed include:
033057: Mar 11 09:33:59.015: RADIUS: Retransmit to (10.0.5.11:1812,1813) for id 1645/127
033058: Mar 11 09:33:59.015: RADIUS: PAC key found while sending request (rctx:0x75EB6A555690)
033059: Mar 11 09:33:59.015: RADIUS: Message Authenticator encoded
033060: Mar 11 09:33:59.015: RADIUS(00000000): Started 4 sec timeout
And in ISE logs:
RadiusListener,2026-06-28 23:21:22,149,WARN ,0x7efea2055700,NIL-CONTEXT,Hit context count limit. Dropping request!,RADIUSListener.cpp:339
The issue was resolved by applying a Cisco-provided hotfix that addresses a deadlock condition in the ISE SessionUpdater process. Follow these steps to implement the resolution:
Download the hotfix files from Cisco and transfer them to your ISE node repository. The hotfix files are:
ise-apply-CSCwv71473_3.5.0.527_patch3-SPA.tar.gz (installation file)
ise-rollback-CSCwv71473_3.5.0.527_patch3-SPA.tar.gz (rollback file)
Install the hotfix using the CLI command:
application install ise-apply-CSCwv71473_3.5.0.527_patch3-SPA.tar.gz <repository name>
Caution: Possible. The details mentioned here appears to contain procedures or commands that could cause significant impact if executed. Please ensure the above procedures or commands have been evaluated by an SME or Business Unit before executing or recommending.
Verify the hotfix installation using one of these commands:
show logging application hotpatch.log
Or:
show version
After installation, monitor the ISE node for a minimum of two weeks to ensure the issue does not recur. The hotfix installation requires a service restart, so plan the installation during a maintenance window.
If needed, the hotfix can be removed using the rollback command:
application install ise-rollback-CSCwv71473_3.5.0.527_patch3-SPA.tar.gz <repository name>
To monitor RADIUS context allocation, use:
show logging application prrt-server.log | include ContextN
Enable runtime AAA debugs before any future troubleshooting:
debug aaa runtime
The root cause was identified as a deadlock condition in the ISE SessionUpdater process. When simultaneous accounting start and stop events occurred for the same session, two threads entered a circular wait scenario:
Thread-162: owned cacheLock, waited for SessionInfo lock
Thread-125: owned SessionInfo lock, waited for cacheLock
This circular wait led to ContextN leaks and a continuously growing RADIUS processing queue. The queue buildup eventually exhausted available RADIUS contexts, causing authentication latency and network devices to mark the ISE node as RADIUS DEAD.
The issue is documented in Cisco bug ID CSCwv71473.
| Revision | Publish Date | Comments |
|---|---|---|
1.0 |
17-Nov-2026
|
Initial Release |