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

Conference vaxaxp::vmsnotes

Title:	VAX and Alpha VMS
Notice:	This is a new VMSnotes, please read note 2.1
Moderator:	VAXAXP::BERNARDO

Created:	Wed Jan 22 1997
Last Modified:	Fri Jun 06 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	703
Total number of notes:	3722

556.0. "pool memory expantion" by AXCL04::LAGIOIA (We do ...the best) Fri May 02 1997 07:33

    Hello!
    Our customer has a vax running openvms v6.2 and has this (known)
    problem:
    
    OJVA01:\PPAHUD>sh mem/pool/full
           System Memory Resources on  2-MAY-1997 10:21:34.49
    
    Nonpaged Dynamic Memory      (Lists + Variable)
    ->Current Size (bytes)      14080000    Current Total Size (pages)  27500
      Initial Size (NPAGEDYN)    3520000    Initial Size (pages)         6875
    ->Maximum Size (NPAGEVIR)   14080000    Maximum Size (pages)        27500
      Free Space (bytes)          268032    Space in Use (bytes)     13811968
      Size of Largest Block         5184    Size of Smallest Block         64
      Number of Free Blocks         1024    Free Blocks LEQU 64 Bytes     634
      Free Blocks on Lookasides       66    Lookaside Space (bytes)     26752
    
    Paged Dynamic Memory
      Current Size (PAGEDYN)     3031552    Current Total Size (pages)   5921
      Free Space (bytes)         2229968    Space in Use (bytes)       801584
      Size of Largest Block      2226656    Size of Smallest Block         16
      Number of Free Blocks          114    Free Blocks LEQU 64 Bytes     111
    
    Doing a delete of listprepop.dat and rebooting the systen it clears the
    problem and after around 6 months the problem occurs again.
    
    After 6 months I think that is normal but the customer wants to know if 
    there is "a fix". 
    
    I just need a confirmation.
    
    Best regards
    Domenico

T.R	Title	User	Personal Name	Date	Lines
556.1		BSS::JILSON	WFH in the Chemung River Valley	`Fri May 02 1997 09:40`	4
	If the customer is running Pathworks then you need to look at the STARS article [OpenVMS,PW-VMS] PATHWORKS V5.0 Periodically Consumes Nonpaged Pool Jilly
556.2	no pathworks	AXCL04::LAGIOIA	We do ...the best	`Fri May 02 1997 09:51`	9
	Hello! Thanks for your quick answer but they do not have pathworks. Any other idea? Tanks Cheers
556.3	IPMT	XDELTA::HOFFMAN	Steve, OpenVMS Engineering	`Fri May 02 1997 11:16`	12
	There are cases where the self-tuning pool support does cause problems, due to the local application loading. There exists the potential for changes in this area and (if memory serves) there are changes being made, but you will want to pursue this through formal channels (such as an IPMT), as someone from engineering will likely need to characterize the application load and look at the self-tuning data to determine what has happened here. The IPMT would likely be relatively low priority, as you appear to have a workaround.
556.4		BSS::JILSON	WFH in the Chemung River Valley	`Fri May 02 1997 11:31`	33
	Nonpaged Dynamic Memory (Lists + Variable) ->Current Size (bytes) 14080000 Current Total Size (pages) 27500 Initial Size (NPAGEDYN) 3520000 Initial Size (pages) 6875 ->Maximum Size (NPAGEVIR) 14080000 Maximum Size (pages) 27500 -> Free Space (bytes) 268032 Space in Use (bytes) 13811968 Size of Largest Block 5184 Size of Smallest Block 64 Number of Free Blocks 1024 Free Blocks LEQU 64 Bytes 634 Free Blocks on Lookasides 66 Lookaside Space (bytes) 26752 Since the pool is in use you will need to take a look at who/what is using the pool. There are pool leaks in a few different pieces of software. If the customer is running DECNET/OSI then make sure they have the latest version with the latest patches. To see where the pool is being consumed you need to look at the packets for the largest % consumer $ ANA/SYSTEM READ/EXEC READ SYS$SYSTEM:SYSDEF SHOW POOL/SUMMARY/NONPAGE Once you see what packet type is the largest consumer then you can do a SHOW POOL/HEAD/TYPE=blah/NONPAGE This will show you the address of the packet, the size of the packet (in decimal) and the 1st 4 longwords in the packet (in hex and in ASCII). If you can't determine anything from the longwords then you'll need to dump some of the packets with an EXAM address;^Dsize The US CSC has some internal tools that can be used to trace who/what is allocating pool and not releasing it. Normally this type of work is a billable service to customers unless it is Digital software casusing the pool expansion. Jilly
556.5		MOVIES::WIDDOWSON	Rod OpenVMS Engineering. Project Rock	`Fri May 02 1997 12:15`	8
	RE: .1,.2.,3 note that .0 says that this is a Known problem, and the question is whether there is a fix to this known problem. RE: 0. What is the reference (STARs for preference) which makes you state that this is known ? It might be worthwhile asking for an update to the IMPT case...
556.6	IPMT	XDELTA::HOFFMAN	Steve, OpenVMS Engineering	`Fri May 02 1997 12:42`	11
	re. 5 : note that .0 says that this is a Known problem, and the question is : whether there is a fix to this known problem. Folks are aware of workloads where the self-tuning mechanism causes problems, and are aware of specific workloads that can cause this. Assuming this is not the Pathworks "storm" problem, and is actually a problem with the memory management self-tuning, the IPMT will get the request into the queue, and will get this particular workload looked at to see if this is a new/different failure in the self-tuning.
556.7	there isn't much left to manage	EVMS::GRANT		`Fri May 02 1997 14:04`	3
	But .4 points out the real problem in this situation. The variable list is short and there isn't much on the lookaside lists. Nearly all of pool is currently allocated. Step 1 is to find out who has it.
556.8		PHXSS1::HEISER	Maranatha!	`Tue May 27 1997 20:15`	10
	\|the pool. There are pool leaks in a few different pieces of software. If \|the customer is running DECNET/OSI then make sure they have the latest \|version with the latest patches. To see where the pool is being consumed \|you need to look at the packets for the largest % consumer fwiw, DECnet/OSI v6.3 ECO6 is the latest and has the memory leak. ECO7 is being worked on. regards, Mike