HP Serviceguard for Linux Version A.11.16 Release Notes, Third Edition, August 2006

Serviceguard for Linux Version A.11.16 Release Notes
Known Problems and Workarounds
Chapter 158
Configuring multiple users with the same UID results
in non-deterministic UID to username resolution.
• What is the problem? When most database schemes are used with
user/password lookup, configuring multiple users with the same UID
results in non-deterministic UID to username resolution. For
example, if you have three UID 0 users with user names “root”,
“jacksu”, and “jillsu” and you login as “root”, invoking the id
command may show your username as “root”, “jacksu”, or “jillsu”.
The problem has been demonstrated with daemon called nscd, which
keeps a cache around and searches through a list of hashes instead of
sequentially through the /etc/passwd file. However, this is a generic
problem and may be exhibited with various sorts of password
database schemes (including NIS, LDAP, or storing /etc/passwd in
a local database using nss_db and libdb).
This may result in Serviceguard for Linux commands that fail on
user lookup and Serviceguard log files that list the incorrect user as
executing a command.
• What is the workaround? Configure only one user with UID 0 and the
name for this user must be “root” on all nodes in a Serviceguard for
Linux cluster. Users with UID 0 on remote (non-cluster) linux nodes
that are configured to execute Serviceguard commands or run
Serviceguard Manager must also follow these rules. All non-root
users on Linux systems that are individually configured to execute
Serviceguard commands or run Serviceguard Manager must have
unique UIDs.
Resource monitor returning -1 second running time
• What is the problem? The resource monitor agent running time is
reported by querying the resource monitor daemon. This value is
refreshed on every polling interval of the agent. Sometimes, it will
report a running time of -1 seconds. This does not mean the agent
ran for -1 seconds. It is an error in the daemon.
• What is the workaround? The reported status is corrected in the next
polling interval.