[Search for users]
[Overall Top Noters]
[List of all Conferences]
[Download this site]
| Title: | NAS Message Queuing Bus |
| Notice: | KITS/DOC, see 4.*; Entering QARs, see 9.1; Register in 10 |
| Moderator: | PAMSRC::MARCUS EN |
|
| Created: | Wed Feb 27 1991 |
| Last Modified: | Fri Jun 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 2898 |
| Total number of notes: | 12363 |
2814.0. "FIFO order with P_DEL_MODE_WF_DQF" by BARNA::DSMAIL () Thu Mar 20 1997 06:24
I have a question about FIFO order. I have a customer with a cluster
of ALPHA 1000 servers running OpenVMS V6.2-1H3 and DECmessageQ V3.2A-
22(3222). There is a process running in one node of the cluster sending
messages with PPM_DELIVERY = PDEL_MODE_WF_DQF and PPM_UMA =
PDEL_UMA_SAF. It works like this, sends a lot of messages, five minutes
of inactivity and sends a lot of messages again.
There is a process running in the other node of the cluster receiving
and confirming this messages and we have seen that this process doesn't
read the messages in order. For example,
Process 1
---------
10:00:00:10 Send message 1
10:00:00:20 Send message 2 these four messages are sent very
10:00:00:35 Send message 3 rapidly, in the same second
10:00:00:47 Send message 4
10:05:00:50 Send message 5
10:05:00:68 Send message 6
Process 2
---------
10:00:10:44 Read message 1
10:05:07:33 Read message 5
And after it reads message 2,3,4 and 6. We don't loose any message but
we don't have it in order. If we stop the process that gets the
messages before 10:05, then the process reads them in order FIFO
(2,3,4,5,6).
Is there any parameter that we have to modify to read them in FIFO
order? IS a possible solution to wait one or two seconds between "puts"
in the same queue?
Thanks in advance
Asun Ariza
DEC Barcelona
| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 2814.1 | The only way I know that you can get this behavior... | KLOVIA::MICHELSEN | BEA/DEC MessageQ Engineering | Thu Mar 20 1997 09:08 | 14 |
| ...is when at least one of following is true:
- Messages are sent with different priorities
- Messages are sent with different delivery modes and/or UMAs
- Messages are read with a non-FIFO selection filter
otherwise it's a bug.
Marty
|
| 2814.2 | | PAMSIC::POPP | Deep in the Heart... | Thu Mar 20 1997 11:01 | 11 |
|
If your messages don't meet any of the criteria outlined in the previous
reply and you believe the problem is with DMQ then I will need you to
provide me with MRS server & header tracing on both the sending and receiving
groups. Tracing is enabled via the DMQ$MGR_UTILITY option RO. Select
server tracing YES and header tracing YES and set the trace output to EVL_LOG.
-Lisa Popp
DmQ Engineering
|
| 2814.3 | Problems obtaining the trace output | BARNA::DSMAIL | | Wed Apr 02 1997 11:38 | 21 |
|
I don't known how to obtain the trace. The sending and receiving group
is the same. So, when i select he RO option from the DMQ$MGR_UTILITY, which
queue must i select? the MRS_SERVER or the message queue?
I have selected the MRS_SERVER queue and then
DMQ$SERVER_TRACE = YES
DMQ$HEADERS = YES
DMQ$TRACE_OUTPUT = EVL_LOG
Must I stop and restart the decmessageq group? Where is the trace
output kept?
Sorry because i have answered you so late, but i'm usually at a
customer and have not access to this conference.
Thanks in advance.
Asun Ariza.
|
| 2814.4 | | PAMSIC::POPP | Deep in the Heart... | Wed Apr 02 1997 16:08 | 13 |
|
> I have selected the MRS_SERVER queue and then
> DMQ$SERVER_TRACE = YES
> DMQ$HEADERS = YES
> DMQ$TRACE_OUTPUT = EVL_LOG
This is correct. The trace data will be directed to the EVL log file
dmq$log:dmq$evl_bbbb_ggggg.log. After you have finished collecting the
trace data then you may go back into the mgr_utility and turn tracing back
off (server_trace=NO, headers=NO). Then use the LC menu to close and open
a new EVL log file. Send me to old one with the EVL tracing.
-Lisa
|