Novell Open Enterprise Server 11 (OES 11) Linux Support Pack 2
An Novell OES Cluster was serving an cluster-enabled NSS volume through AFP. Customer found that when attempting "rcafptcpd stop" (to stop the service), the stop attempt would delay a few seconds and then fail.
Reports from "ps" confirmed the process was still running. Only kill -9 would successfully stop it.
Customer was running OES 11 SP2 without any maintenance updates applied. After applying maintenance updates through the March 2015 maintenance release, the problem disappeared.
Not much is known about this case. The exact cause of the problem was not confirmed, not the nature of the solution within the maintenance updates. It is not known if the presence of Novell Clustering was influencing this.
As a side note, it is not advisable to stop afptcpd manually if clustered resources are being shared with AFP. Resources should be migrated or offlined first.