On Thu, 2009-03-26 at 03:43 +0000, Ganesha, Raghavendra Pandimakki wrote: > Hi Shuah, > > The hotswap events, except surprise extraction should have event severity set > to INFORMATIONAL. > > This is according to the SAF HPI B.03 spec, section 7.9.1 (page 222). > "Hotswap events indicating normal state transitions are issued with a > severity of SAHPI_INFORMATIONAL. Events that indicate a "surprise extraction" > are issued with the severity set in the ResourceSeverity field of the RPT > entry for the resource." > > This change was done as part of fix to bug #2529314. > > Regards, > PG
Hi PG, Thanks for taking the time to give details on the change made in this area. The following is the text from this section: "Hot swap events indicating normal state transitions are issued with a severity of SAHPI_INFORMATIONAL. Events that indicate a “surprise extraction” are issued with the severity set in the ResourceSeverity field of the RPT entry for the resource. A “surprise extraction” is a transition to the NOT PRESENT state from any state other than INACTIVE. HPI implementations may generate hot swap events that re-announce the current hot swap state of a resource. These events should have matching values for the previous and current hot swap states, should be issued with a severity of SAHPI_INFORMATIONAL and should have a cause of state change of SAHPI_HS_CAUSE_UNKNOWN." The use of "are issued with a severity of SAHPI_INFORMATIONAL." to me says, these events are at SAHPI_INFORMATIONAL, not should be. So using a higher severity should be ok. "Should" is used later on in the context of generating re-announce events. The emphasis is on maintaining the same severity level for re-announced events. The way I read this is that HPI implementations can choose a severity level of SAF_INFRMATIONAL or higher for HS events and maintain the same level for re-announce as well. I am going by the use of "are" in the very first sentence. Also some of these events are a result of a blade power down and if a blade powers down because of some failure or overheating, I would think it should be an event at a higher severity than informational. I plan to discuss this in one of the upcoming SAF-HPI TWG meetings. Now the openhpi infrastructure. Infrastructure code by default logs events that are MINOR or higher to EL and drops the rest. HS events are important and they shouldn't be dropped. Changing infrastructure to log informational events will flood the EL. Customers need to know to set the global parameter to log informational events. So for now, making these events critical (same as the way back 2.12) is reasonable choice and get clarification from the TWG. Hope this helps. Thanks again for taking the time to send email explaining the changes made. -- Shuah ------------------------------------------------------------------------------ _______________________________________________ Openhpi-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/openhpi-devel
