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

Conference hydra::axp-developer

Title:	Alpha Developer Support
Notice:	[email protected], 800-332-4786
Moderator:	HYDRA::SYSTEM

Created:	Mon Jun 06 1994
Last Modified:	Fri Jun 06 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	3722
Total number of notes:	11359

3428.0. "Scandinavian Softline Technology - Point 24664" by RDGENG::ASAP () Wed Apr 02 1997 10:07

    Company Name :  Scandinavian Softline Technology - Point 24664
    Contact Name :  Kari KAILAMAKI
    Phone        :  358-9-54956207
    Fax          :  358-9-5124629
    Email        :  [email protected]
    Date/Time in :   2-APR-1997 15:07:19
    Entered by   :  Pauline Snyder
    SPE center   :  REO

    Category     :  UNIX
    OS Version   :  3.2
    System H/W   :  Alpha


    Brief Description of Problem:
    -----------------------------

Our server software uses standard SVR4 message queue facility for IPC.
Now we have serious problems with OSF1 v3.2. 

It seems that when the machine gets heavily loaded messages are sometimes
duplicated by the kernel.

Messages are put on the queue with msgsnd(...) and get from the queue
with msgrcv(...) system calls. Both operations are done in the IPC_NOWAIT
mode.

Now I have carefully traced our programs and I am sure that the message
is always put only once in the queue but still sometimes two successive
msgrcv()s returns the very same message. This happens only when the
machine is very loaded.

Our client having these problems have (uname -a)
   OSF1 www.mobilkom.at V3.2 41 alpha

T.R	Title	User	Personal Name	Date	Lines
3428.1	more info please	RDGENG::SNYDER_P	The speed of time is best set to 1sec/sec	`Tue Apr 08 1997 11:06`	58
	To: [email protected] Cc: [email protected] Subject: Re: Digital ASAP call 24664 - Scandinavian Softline Technology Date: Thu, 03 Apr 97 14:53:36 +0100 From: "Pauline Snyder" <pauline> X-Mts: smtp ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ From: Pauline Snyder Loc: REO/F8 DTN: 830 4096 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Kari, I have looked through our bugs and patches system for old releases and found nothing relevant. As msgrcv and msgsnd are fundamental to a lot of commercial packages it is unlikely that a bug of this magnitude has gone undetected since the Digital UNIX 3.2c release in August '95, so I think we must next look at your software. If you are using a release earlier than 3.2c then you should in any case be looking to upgrade as soon as you can. Is it possible that you are failing to detect an E2BIG error (message is too big for the supplied buffer) and then looking at the previous message still left in the buffer ? In this case the message will be left on the queue. Then maybe next time you do a msgrcv() you supply a buffer of a different size. I have seen this in the past on loaded systems where the message pool handling code was not checking the size of buffer returned was the same as requested. You can check this in 2 ways - either set MSG_NOERROR which will cause a partial message to be received and hopefully a different error condition to arise. Or you can add a bit of code to invalidate the first few bytes of the message after processing. If this does not fit your way of using msgsnd/rcv then we need to look a bit deeper. In this case can you e-mail me directly the following:- output from 'head /etc/motd' the C modules containing the msgsnd/rcv calls An overview of how you are using msgsnd/rcv and the message buffer allocation mechanism. As this only happens when the system is loaded I would guess you cannot easily reproduce it with a cut-down version of the code, but if you can that would be ideal. Best regards, Pauline ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Pauline Snyder Software Partner Engineering, Reading Digital Equipment Company Ltd. Digital Park, Worton Grange Imperial Way email: [email protected] Reading RG2 0TE tel: +44.118.920.4096 England fax: +44.118.920.4146 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ =======================================================================
3428.2	no response - closing	RDGENG::SNYDER_P	The speed of time is best set to 1sec/sec	`Tue Apr 08 1997 11:07`	31
	To: [email protected] Cc: [email protected] Subject: Re: Digital ASAP call #24664 - Scandinavian Softline Technology Date: Tue, 08 Apr 97 15:00:30 +0100 From: "Pauline Snyder" <pauline> X-Mts: smtp ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ From: Pauline Snyder Loc: REO/F8 DTN: 830 4096 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Kari, As I have had no respone to my earlier mail I will close this call now unless I hear from you soon. If you have any further problems please mail the ASAP call desk on [email protected]. Best regards, Pauline ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Pauline Snyder Software Partner Engineering, Reading Digital Equipment Company Ltd. Digital Park, Worton Grange Imperial Way email: [email protected] Reading RG2 0TE tel: +44.118.920.4096 England fax: +44.118.920.4146 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ =======================================================================