[Linux-cluster] Tibco EMS on GFS2

[Laszlo Beres]

Hi Steven,

On Mon, Feb 13, 2012 at 10:38 AM, Steven Whitehouse <swhiteho at redhat.com> wrote:

> I'd be very interested to know what has not worked for you. Please open
> a ticket with our support team if you are a Red Hat customer. We are
> keen to ensure that people don't run into such issues, but we'll need a
> bit more information in order to investigate,

Well, we're still in the investigation phase, so as long as I don't
have direct evidences in my hand I don't want to bother Red Hat
support. Let me briefly summarize our case.

I set up a two noded cluster both on RHEL 5.7 x86_64 on HP DL380 G7
(96 GB RAM, 4 Intel Xeon X5675 / 6 cores) with EMC VMAX storage (on
dm-multipath). We launched our Tibco service in December which has
been working so far without any issues. A few days ago our customers
reported that the service is "slow" - as a system engineer I could not
really do anything with such a statement, so our application guys
deployed a small tool which sends messages to the message bus, and
also measures the message delivery delay. Most of the times it's
around 1-2 ms, but surprisingly found extra high values, which have
affect on dependant applications:

2012-02-08 22:12:31,919 INFO  [Main] Message sent in: 1ms
2012-02-08 22:12:33,974 ERROR [Main] Message sent in: 1053ms
2012-02-08 22:12:34,978 INFO  [Main] Message sent in: 2ms
2012-02-08 22:12:35,980 INFO  [Main] Message sent in: 1ms

At the same time iostat show high utilization:

2012-02-08 22:12:33 avg-cpu:  %user   %nice %system %iowait  %steal   %idle
2012-02-08 22:12:33 0,12    0,00    0,37    4,00    0,00   95,51
2012-02-08 22:12:33
2012-02-08 22:12:33 Device:         rrqm/s   wrqm/s   r/s   w/s
rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
2012-02-08 22:12:33 sdb               0,00     0,00  0,00  0,00
0,00     0,00     0,00     0,00    0,00   0,00   0,00
2012-02-08 22:12:33 sde               0,00     0,00  0,00  0,00
0,00     0,00     0,00     0,00    0,00   0,00   0,00
2012-02-08 22:12:33 sdh               0,00     0,00  0,00  0,00
0,00     0,00     0,00     0,00    0,00   0,00   0,00
2012-02-08 22:12:33 sdk               0,00     0,00  0,00  0,00
0,00     0,00     0,00     0,00    0,00   0,00   0,00
2012-02-08 22:12:33 sdn               0,00     0,00  0,00  0,00
0,00     0,00     0,00     0,00    0,00   0,00   0,00
2012-02-08 22:12:33 sdq               0,00     0,00  0,00  0,00
0,00     0,00     0,00     0,00    0,00   0,00   0,00
2012-02-08 22:12:33 sdt               0,00     0,00 56,00  0,00
0,34     0,00    12,43     0,12    2,21   2,20  12,30
2012-02-08 22:12:33 sdw               2,00    39,00 337,00 28,00
1,84     0,12    10,98     2,40    2,75   2,34  85,40
2012-02-08 22:12:33 dm-5              0,00     0,00 394,00 201,00
2,18     0,79    10,19     2,98    1,94   1,65  97,90

>From our service point of view these are off-peak times, less messages
were directed to EMS.

All unimportant services are disabled, no jobs are scheduled in system.

Also the occurence pattern is quite strange:

2012-02-10 03:01:24,333 ERROR [Main] Message sent in: 2306ms
2012-02-10 03:02:04,953 ERROR [Main] Message sent in: 1221ms
2012-02-10 03:11:29,725 ERROR [Main] Message sent in: 1096ms
2012-02-10 03:31:35,195 ERROR [Main] Message sent in: 1051ms
2012-02-10 03:36:36,943 ERROR [Main] Message sent in: 1263ms
2012-02-10 04:01:17,059 ERROR [Main] Message sent in: 1585ms
2012-02-10 04:01:41,953 ERROR [Main] Message sent in: 1790ms
2012-02-10 04:02:42,953 ERROR [Main] Message sent in: 1307ms
2012-02-10 04:06:03,181 ERROR [Main] Message sent in: 2305ms
2012-02-10 04:08:48,844 ERROR [Main] Message sent in: 1294ms
2012-02-10 04:12:52,282 ERROR [Main] Message sent in: 1350ms
2012-02-10 05:01:00,411 ERROR [Main] Message sent in: 1143ms
2012-02-10 06:01:01,550 ERROR [Main] Message sent in: 1291ms
2012-02-10 06:01:48,957 ERROR [Main] Message sent in: 1092ms


Regards,

Laszlo