[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference specxn::decevent

Title:	DECevent -- keywords KIT* and PAK are available
Notice:	DECevent V2.2 available - SHOW KEY/FULL to find note numbers
Moderator:	DSNENG::KOLBEN

Created:	Fri Jan 07 1994
Last Modified:	Thu Jun 05 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	307
Total number of notes:	1463

296.0. "DIAG/ANAL V2.4 error - status code 10450592" by KERNEL::SCOTT (You can trust a Teddy Bear!) Wed May 07 1997 11:56

    Hi,
    
    	Can someone please decode this error from DECevent V2.4 on ALPHA VMS
    V6.2:- 	{and teach me how I could do this myself?}
    
    
    PINKY$ diag/analyze/since=today
    
    DECevent V2.4
    
    {copyright etc left out for brevity}
    
    _DIAGNOSE-FAT: Generate analysis method failed
    _DIAGNOSE-FAT: Generate analysis method failed
    _DIAGNOSE-FAT: No message found for status code 10450592 from
    ana_eng__f_notify in thread 7
      ...Offending facility has no translation
      ...facility number: 2
      ...status code: 257748
    _DIAGNOSE-FAT: calloc RTL call failed.
    _DIAGNOSE-ERR: not enough core,
      from svc_knl__open_library in thread 7
    
    _DIAGNOSE-FAT: Execution failed
    
    
    I ran the command procedure indicated in 122.1 and that didn't show any
    process quotas being exceeded (or even getting close to limits). There
    are no pool problems and the pagefile has plenty of free space. I've
    tried reducing the errlog to 10 blocks and still no joy. 
    
    
    regards  Roland Scott

T.R	Title	User	Personal Name	Date	Lines
296.1		OLD1S::SYSTEM	PM&D PSE Tools Support	`Wed May 07 1997 13:14`	6
	It appear you have run across a problem we have been working to resolve. If the manual analysis find zero entries for a supported device the tool crashes. We are working on this problem. It has been hard to reproduce. Keith
296.2		KERNEL::SCOTT	You can trust a Teddy Bear!	`Wed May 07 1997 13:56`	8
	Hi Keith, Thanks for the reply. Does this mean that if there is something for analyze to get its teeth into it won't crash? Would it then follow that a mail would be sent and a call logged if SICL is enabled when there is something to report? roland
296.3		OLD1S::SYSTEM	PM&D PSE Tools Support	`Wed May 07 1997 14:19`	20
	Roland, What gets reported via the automatic process is contained the in prm state db. A entry for each event for a device covered by analysis is written to that database. When a device triggers automatic analysis, the process goes to the perm state db and pull all entries for that device. The automatic process does not go to the error log for entries on supported devices. For devices not supported by analysis. The process will format the sicl message and the customer profile and add the error log information for that device and ship it to the CSC via DSNlink. We ship up to 60 blocks of information with this type of call. The 60 block limit is set by DSNlink. Manual analysis uses the error log file and not the perm state db. HTH Keith
296.4		KERNEL::SCOTT	You can trust a Teddy Bear!	`Thu May 08 1997 06:12`	21
	Thanks Keith, I was reading the help on DIAG/ANALYZE and it says:- DIAGNOSE /ANALYZE Provides the capability to either manually or automatically analyze entities on an operating system. Notification to individuals on selected mailing lists, as to the results of the analysis, is also supported. It seemed to me that the automatic analysis would use the same code or logic for automatic calls, regardless of where the error information came from. Does this "perm state db" work in the same way as VAXSIM$CLUSTER.DAT ? thanks again, roland
296.5		OLD1S::SYSTEM	PM&D PSE Tools Support	`Thu May 08 1997 13:02`	17
	Roland, From what I can pull out of the vaxsim cobweb seems to recall that the cluster .dat contained info with relation to the cluster monitor. The PRM state db contains the error log info generated by the supported devices. With any luck, we will have vaxsim like support for cluster environments one of these days. Until we can figure out KNL locking cluster support is out. Two processes open the knl and it becomes corrupt. HTH Keith
296.6		KERNEL::LOANE	Comfortably numb!!	`Thu May 08 1997 18:18`	23
	Keith, VAXSIM$CLUSTER.DAT contains a whole bunch of data (Margins, Checkpoints, Triggers) as well as device records. Each record is based on a particular device class of error and is timestamped (1's complement of the most significant longword of system time) so that when VAXsimPlus initialises, it can build the appropriate in-memory data structures with valid, up to date information about the current state of error rates.....it doesn't go anywhere near the errorlog. Hence, if a newly booted system had a device record containing 15 events within the last 24 hours .AND. the margin was 15, then a subsequent event (still within the 24 hour period) would trigger the fault manager. It's only during the fault manager phase that we'd read the errorlog. I believe that's what Roland was alluding to. Cheers Chris P.S. I just remembered that VAXsimPlus DOES read the errorlog at initialisation...but purely to pull any ELE information that may have been loaded from a dump file (i.e. using the Checkpoint records).
296.7		OLD1S::SYSTEM	PM&D PSE Tools Support	`Thu May 08 1997 19:27`	13
	Chris, From what I remember once a trigger occurs. the Tool does a merge of all error logs in the cluster and extract the data from all system with relation to the device. The cluster .dat files man function was to contain the monitor information. It kept track of all the device it found during start up. Any event that occured would be written to that device's locations. This allowed you to go back x number of days using the /since switch. Boy it's been a long time since that "Work of Art" has been discussed. Keith