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

Conference noted::pwv50ift

Title:Kit: Note 4229; Please use NOTED::PWDOSWIN5 for V4.x server
Notice:Kit: Note 4229; Please use NOTED::PWDOSWIN5 for V4.x server
Moderator:CPEEDY::KENNEDY
Created:Fri Dec 18 1992
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:4319
Total number of notes:18478

4261.0. "PWMON LMSRV_QUEUE_CHQ, LMSRV RWPAG" by VMSNET::ALLERTON (Episode d'Azur) Wed Apr 16 1997 11:42

    Have a customer at V5.0D ECO3 (with "Sharefile" patched LMSRV image),
    AXP 7000/740 (the only PW node in a cluster) that yesterday experienced 
    performance slowdown on server, and "server unavailable" messages 
    generated at clients that were trying to establish fresh sessions.
    
    Nothing unusual turned up in log files, aside from around 15 successive
    "Server allocating 10 more queue elements" lines in the server log.  
    Line counters clean, no pool expansion, though customer believes 
    PWRK$LMSRV process was RWPAG (which from what I've located indicates
    paged pool deficiency, real or understood)
    
    At the time of the slowdown, customer ran PWMON, and noticed that
    LMSRV_QUEUE_CHQ had Curr Waitrs at 121, peak Waitrs at 300,
    Ave. time for last 10 was 3-4 seconds.  LMSRV_QUEUE_SMBQ had Cur waitrs
    between 30-60.  Assuming that the values for LMSRV_QUEUE_CHQ are
    inordinate, we're seeking an explanation of its meaning (reserved
    channels maybe?) and what approaches (necessary steps to evaluate, and
    possible PWRK.INI modifications) might be advised.
    
    Thanks again,
    
    Steve
T.RTitleUserPersonal
Name		DateLines
4261.1UTRTSC::SWEEPI want a lolly...Mon Apr 21 1997 06:493
    Increase page file.
    
    Adrie
4261.2What's being queued?VMSNET::P_NUNEZMon Apr 21 1997 09:325
    Adrie,
    
    Does the CH in LMSRV_QUEUE_CHQ stand for CHannel (if not, what)?
    
    Paul
4261.3UTRTSC::SWEEPI want a lolly...Tue Apr 22 1997 04:429
    Connection Handoff...
    
    A connection request arrives at the listener thread
    in the MCP. The MCP queues this request to the connection
    handoff queue. Now the connection handoff thread in the
    LMSRV gets scheduled, it takes the package off the CHQ and
    processes the new connection by giving it a session context.
    
    Adrie
4261.4UTRTSC::SWEEPI want a lolly...Tue Apr 22 1997 04:4513
    Some additional explanation OF WHAT MIGHT HAVE happened.
    
    New session requests arrive, process needs to page but
    no room in pagefile, so process stalls at VMS level
    waiting for pagefile space to become available.
    
    New sessions requests require session context, which
    is memory in process space. Also its possible that 1 or
    more new worker threads are created that need thread context
    and stack (also process memory). So when the process needs
    more memory and its working set is full, it starts to page.
    
    Adrie
4261.5VMSNET::P_NUNEZTue Apr 22 1997 09:5810
    
    Thanks for the explanation - it really helps us better understand
    the way things are suppose to work.  
    
    Along those lines, if a customer determined the lmsrv process is paging
    too much, they would need to increase the working set size for the
    process, right?  To do so the supported way, means increasing the
    number of clients?  Or some other method?
    
    Paul
4261.6UTRTSC::SWEEPI want a lolly...Wed Apr 23 1997 04:0614
    :Along those lines, if a customer determined the lmsrv process is paging
    :too much, they would need to increase the working set size for the
    :process, right?  
    
    Right
    
    :To do so the supported way, means increasing the
    :number of clients?  Or some other method?
    
    Easiest is to increase the nr of clients. Then the configurator
    will do all calculations for you, and check with autogen if all
    system params are all set.
    
    Adrie
4261.7UTRTSC::SWEEPI want a lolly...Wed Apr 23 1997 04:076
    The configurator doesn't check for the pagefile
    being large enough. So this is something you must
    do be hand. (I think, I am not sure if autogen keeps
    track without the feedback qualifier).
    
    Adrie
4261.8How?VMSNET::P_NUNEZWed Apr 23 1997 11:0117
    
    Adrie,
    
    Re. .1 
    
    Do you mean the process pagefile quota (which may necessitate
    an increase in a system pagefile) or the system pagefile?
    
    re. .7
    
    Do we increase the pwrk$lm processes pagefile quota in pwrk.ini:
    
    [Pgflquo]
    
    lmsrv=<new value>
    
    Paul
4261.9UTRTSC::SWEEPI want a lolly...Thu Apr 24 1997 06:1210
    Not sure
    
    I would first increase the size of the system pagefile
    and if that doesn't help raise the pgflquota.
    
    I thing that RWPAG can only happen when the pagefile
    is full-Resource wait waiting for pagefile space, but
    to be sure consult your vms specialist.
    
    Adrie