Votos - 0, Puntuación media: 0
(
)
)
|
Otros manuales para este modelo:
Resumen del manual
HP Fabric Clustering System
Known Problems and Issues
Transition from Point-to-Point to Switch-Based Fabric Restriction
To move an HP Fabric Clustering System server from a point-to-point to a switch-based fabric, or vice versa,
the server must be rebooted for the fabric clustering to function correctly.
NOTE
The server must be rebooted after the HP Fabric Clustering System cables are disconnected
from the old fabric.
MPI and PA Systems
MPI is not a supported HP Fabric Clustering System protocol for PA systems.
Suspending an HP Fabric Clustering System HCA Using the olrad (1M) Command
Suspending an HP Fabric Clustering System HCA using the olrad (1M) command requires the forced option
-f to be used. The presence of IPoIB interfaces (for example, lan9000) on the HCA results in the
CRA_DATA_CRITICAL error during the critical resource analysis on the HCA, and thus fails the olrad
suspend operation.
NOTE
Use of the forced option -f will kill any user space applications that are using the affected
HCA.
The recommended sequence of actions is:
1. Run olrad -c [slot_id] to launch critical resource analysis on the HCA.
2. Review the command output and look for any user space applications that are using the affected HCA.
3. If there are no user space applications and if the olrad command still returns the CRA_DATA_CRITICAL
error, perform olrad -f-r [slot_id] to force the suspend operation on the HCA.
4. After replacing the defective card, resume the driver instance with olrad -R [slot_id].
Potential for Stale Path Information
An HP Fabric Clustering System server (Server A) can obtain stale path information for another HP Fabric
Clustering System server (Server B) if any of the following conditions are true for Server B:
•
Server B has been rebooted.
•
One or more HCAs on Server B have been reset.
•
One or more HCAs have been part of an OL* replace operation on Server B.
Stale path information for Server B means that Server A has stored an entry in its address cache that no
longer resolves to the correct address for Server B. This stale path information can be present on Server A
from 0 to a maximum of 10 minutes. This can result in loss of connectivity for a maximum of 10 minutes.
Either of the following actions can be taken to work around this restriction:
•
Us e th e /opt/rdma/bin/itutil -f command to flush path information from the server containing the
stale path information (Server A).
•
Wait for 10 minutes and retry. After 10 minutes, the stale path information will have been deleted from
Server A.
8
Chapter
...Otros modelos de este manual:
Equipo de red - HP-UX 11i v2 I/O Cards (141.36 kb)