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

Conference help::dns

Title:DECdns - Digital's Distributed Name Service
Notice:not to be confused with DNS: Domain Name Service (Kits: 1420,947))
Moderator:BULEAN::WHEATER
Created:Tue Apr 14 1987
Last Modified:Wed Jun 04 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:1443
Total number of notes:5833

1429.0. "Request Pool Exhausted" by YAKKA::KINGSMILL (Geoff Kingsmill, Australia) Wed Feb 19 1997 06:19

VMS/VAX V6.2
DECDNS ECO 70

Twice now on different systems I have seen the DNS hang with hundreds of the
following messages in dns$server.log

Tue Feb 18 23:00:01.209: W Request pool exhausted. Port shutdown.

I monitored the dns$server process quotas and could not find any quota 
anywhere near exhausted. Pgflquota still had 50% (100000 pages) unused.
WSSIZE was still way below WSEXTENT/WSMAX.

I also checked that CTLPAGES was less than 128 pages greater than CTLIMGLIM
as per John Weir's decnet-osi note 3762. 

Unfortunately I only had a short time to look at this as it was severely 
affecting the customers operation. I didn't get to look at other DECnet 
process's such as NET$ACP.

What resource is "W Request pool" drawing from? 
Any ideas?

Thanks,
Geoff..
T.RTitleUserPersonal
Name
DateLines
1429.1adjust your server configurationDRAGNS::WHEATERWed Feb 19 1997 13:0732
    Request pool is an internal resource used by the DNS server.

    I looks like you need to modify your configuration in the
    sys$sysdevice:[dns$server]dns.conf file

    An example of what the easynet uses is shown below:

    dns.dnsd.idle_conn_timeout: 60
    dns.dnsd.null_port_timeout: 60
    dns.dnsd.maximum_handlers_quota: 1000
    dns.dnsd.maximum_buffers_quota: 1000
    dns.dnsd.ta_conn_quota: 400
    dns.dnsd.back_conn_quota: 100

    Your workload may not require the quotas to be as high as easynet but
    apparently they have to be higher than the default values.

    You can use the following commands in DNS$DIAG to monitor quota
    utilization:

    request handlers, buffers :   sum rpool
    ta_connect_quota          :   sum appl [ta_receiver_id from ex nsb command]
    back_connect_quota        :   sum appl [back_receiver_id from ex nsb ]

    It would be helpful to know how may ports you typically have open. We
    are considering increasing the default values for our next ECO.

    regards,
    Bob Wheater
    DECdns Engineering

1429.2dns$diag help requiredYAKKA::KINGSMILLGeoff Kingsmill, AustraliaWed Feb 19 1997 22:1240
Bob,
   Thanks for the prompt and detailed response.

<    You can use the following commands in DNS$DIAG to monitor quota
<    utilization:
<
<
<    request handlers, buffers :   sum rpool

$ mcr dns$diag sum rpool command at present shows the following:-

                Reserved  Available  Standby  Maximum
Handlers:              2          8       10      100
Buffers:               2         18       20      200

Does this mean that their are only 92 handlers free and 82 buffers free? If so
then this would be too low? Correct?

< You can use the following commands in DNS$DIAG to monitor quota
<    utilization:
<
<    ta_connect_quota          :   sum appl [ta_receiver_id from ex nsb command]
<    back_connect_quota        :   sum appl [back_receiver_id from ex nsb ]

$ mcr dns$diag sum appl [ta_receiver_id from ex nsb command]
Parser error. Expecting address constant. 'appl' isn't one
$ mcr dns$diag sum appl sum appl [back_receiver_id from ex nsb ]
Parser error. Expecting address constant. 'appl' isn't one

Is this the right command or am I meant to do some substitution here?

< It would be helpful to know how may ports you typically have open. We
<    are considering increasing the default values for our next ECO.

Is this simply the number of DECnet links to the dns$server object or is there 
a dns command to show the number of ports?

Thanks,
Geoff..

1429.3use help commandDRAGNS::WHEATERThu Feb 20 1997 12:3561
<    You can use the following commands in DNS$DIAG to monitor quota
<    utilization:
<
<
<    request handlers, buffers :   sum rpool

$ mcr dns$diag sum rpool command at present shows the following:-

                Reserved  Available  Standby  Maximum
Handlers:              2          8       10      100
Buffers:               2         18       20      200

Does this mean that their are only 92 handlers free and 82 buffers free? If so
then this would be too low? Correct?
### No it does not. 98 free handlers, 198 free buffers. Reserved indicates
### that they are in use. Available means that they have been used and are
### ready to be reused, Standby is how may to keep around before returning
### to pool. Maximum is the largest number that can be reserved
### at one time (in use at one time)
###
### It looks like these values are still set to the default and the
### the server doesn't have much activity on it.

< You can use the following commands in DNS$DIAG to monitor quota
<    utilization:
<
<    ta_connect_quota          :   sum appl [ta_receiver_id from ex nsb command]
<    back_connect_quota        :   sum appl [back_receiver_id from ex nsb ]

$ mcr dns$diag sum appl [ta_receiver_id from ex nsb command]
Parser error. Expecting address constant. 'appl' isn't one
$ mcr dns$diag sum appl sum appl [back_receiver_id from ex nsb ]
Parser error. Expecting address constant. 'appl' isn't one

### NO! you were supposed to issue two dns$diag commands
###
### 1. ex nsb
### 2. take the address shown in ta_receiver_id and plug in into the
###    the next command
### 3. ex appl ta_receiver_id_address
###
### the diag program has a help command, I suggest that use it to show
### you the format of these commands.

Is this the right command or am I meant to do some substitution here?

< It would be helpful to know how may ports you typically have open. We
<    are considering increasing the default values for our next ECO.

Is this simply the number of DECnet links to the dns$server object or is there 
a dns command to show the number of ports?

### No!, use the dns$diag command: sum appl ta_receiver_id
### it will give current and Peak ports information

Thanks,
Geoff..



1429.4not everything is in helpYAKKA::KINGSMILLGeoff Kingsmill, AustraliaFri Feb 21 1997 02:298
Bob,
   Thanks. Sorry I misinterpreted your previous commands. They were quite
straight forward now that I read it again. There is currently plenty of spare
pool and ports. I'll monitor this and see if they grow over time.

Thanks,
Geoff..

1429.4YAKKA::KINGSMILLGeoff Kingsmill, AustraliaFri Feb 21 1997 02:497
Bob,
   How embarrassing. I missed the obvious. There is currently plenty of spare
pool and ports. I'll monitor this and see if they grow over time.

Thanks,
Geoff..