Procurve cacti graphs show steady discards out

[fcocquyt]

Hi,
I am a sysadmin using cacti to graph and monitor our Procurve v I.10.43
ProCurve J4903A Switch 2824 and
ProCurve J4904A Switch 2848

I'm getting a lot of steady discard out rates from switches that are not pushing more than 200Mb on those ports

What should I do?
We also have associated app and DB issues we fear could be related.

thanks

[Matt Hobbs]

Try enabling qos-passthrough-mode on the switch, this should help with any Gig>100 buffering issues that would be the likely cause of these ifoutdiscards.

[fcocquyt]

We enabled it (defaults to optimized)
We are observing little or no effect on the discards so far...
what about "balanced" "one-queue" modes?
Would those be better ?

Also is a reboot necessary for
Procurve v I.10.43
ProCurve J4903A Switch 2824 and
ProCurve J4904A Switch 2848?

The 3600 doc says no reboot req'd - we rebooted our 28xx's

thanks

[Matt Hobbs]

I didn't read this correctly the first time, seems the ports that you're seeing the discards on are Gigabit ports? In that case, there should be no issue unless they're being oversubcribed at brief intervals. You could try the one-queue option, also make sure you update to the latest I.10.67 at the same time:

http://www.hp.com/rnd/software/J49031067.htm