View Full Version : Provider-1 Running Slow

2006-04-19, 12:16
I have Provider-1 NGX on SPLAT and out of the blue the CLI is running slow (very slow). The MDS also shows as disconnected in the P1GUI. However all the CMAs work fine on it and I am able to connect top them via Dashboard with no issues.

Anyone have any ideas?

2006-06-01, 13:21
I have P-1 NGX R60A running on Solaris. We run in an HA environment.
Most of our systems run 8 GB of ram, somehow, one of our systems ended up with 2 gb. Periodically, especially when some "heavy" burst of activity hits, the server becomes slow to log into (ssh or telnet), the MDS shows disconnected from our other P-1 systems, we can't connect to it.... but it's not down.

SAR shows I/O wait is comsuming most of the time, little or no idle time.
Top shows similar:
CPU states: 1.5% idle, 2.2% user, 2.8% kernel, 93.5% iowait, 0.0% swap
We have ~ 75 cma's defined on this system.

It appears that perhaps we have some activity that is forcing us into extremely heavy swapping or other i/o activity.
(I personally thought it to be a hardware problem around our disks or disk controller, but, after a firmware upgrade, our Unix team says no prob, must be the app)

So, I might suggest checking amount of RAM, and having at least 2x swap space defined, 3x or 4x not bad. If you run out of swap, your system will do all kinds of things, tasks and processes just don't complete, without much indication why.

Let me know what you find.

2007-04-16, 08:53
Looks like trashing (havily swapping by lack of RAM).
swap isn't usable for such conditions, and disable your swap to avoid the situation (and let the application crash in a lack of memory state, or it may recover.