HP Management Base Installation and User's Guide (May 2009)
hpipmid(8)
NAME
hpipmid -- HP IPMI Daemon
Synopsis
hpipmid [-d device ] [-f ] [-h] [-s LOG] [-v]
Description
hpipmid is started by /etc/init.d/hpmgmtbase. An attempt to start another copy will fail
(with the exception of the -s option). hpipmid talks directly to the system BMC and performs
several tasks:
• Caches the content of the IPMI SEL (System Event Log) and FPL (Forward Progress Log).
Signals are emitted on the System DBus when SEL events are added. These signals allow
other tools (such as hprasd) to be event-driven rather than polled.
• Polls and caches sensor readings on demand.
• Monitors the IPMI SEL (System Event Log) and clears it when it gets too full (80% of capacity).
• Marshals all access to actual BMC from HP manageability tools. This allows cached access
to SEL, FPL, sensors, and invariant data.
• Drives the Virtual Front Panel (VFP) heartbeat indicator in the MP display.
OPTIONS
-d device
Use the argument as the device file. More details can be seen in the hpbmc man page.
-f
Run in the foreground (don't daemonize). Not useful in production.
-h
Suppress the VFPheartbeat; this stops using the IPMI watchdog facility.
-s LOG
Prints the status of a log cache, where LOG must be one of SEL, FPL, or IML. SEL is supported
on all platforms. FPL is only supported on Integrity servers while IML is only supported on
ProLiants
-v
Verbose output; more "v"s gets more output. Only useful with -f.
Files
/var/log/hp/SEL.<platform>.<serial number>
SEL cache backing store.
/var/log/hp/SEL.<platform>.<serial number>
FPL cache backing store.
hpipmid also maintains a System V message queue for use with HP clients. This is the caching
service channel.
WARNING
hpipmid yields a tremendous speed advantage when it is up and running. For example, running
hpbmc SELprintwhen the SEL has 40,000 entries can take six or seven minutes without
hpipmid. This size is not uncommon on high-end Integrity servers. With hpipmid running,
that same operation takes about five seconds.
This assumes the cache has been fully backfilled. Of course, that initial backfill will require those
multiple minutes. While the cache is cache filling, there is no "first" SEL record so operations
that need it will fail. The tail end of the SEL (where new events are placed) is always available.
27