Integrity Servers
cancel
Showing results for 
Search instead for 
Did you mean: 

Rx8620 MP console hung

 
SOLVED
Go to solution
Truxpin
Occasional Visitor

Rx8620 MP console hung

We are running two RX8620 servers.
They are both providing service.
One of the two server has since a few days lost its MP console.
It is not presenting a MAC to the switch (or whatever IP device you connect it to - no matter which speed/duplex cable).
It is also not responding to the Local Serial connection.
To make sure I was not messing up with rs232 cables or settings, I used the same gear/config for the other RX8620 which is instead talking happily on serial and IP...
I tried the MP Reset button in a variety of ways to no avail!
How can I reset the MP console WITHOUT powering off the whole server?
What is it for the little button beside the ATTN and PWR LEDs on the top part of the MP Core I/O card?
Thanks...
I'm going to open an HW case if necessary, but I'm afraid they'll ask me for a power off/on - which nobody here will like at all!
7 REPLIES 7
Murat SULUHAN
Honored Contributor

Re: Rx8620 MP console hung

Hi Truxpin

Do you have configured vPar on the server?

Best Regards
Murat
Murat Suluhan
Truxpin
Occasional Visitor

Re: Rx8620 MP console hung

Hi Murat,
yes I've four running vPars and three of them are serving our business...
Is there a way to communicated to the MP console instance from the vPars?

Thanks!
Stefan Stechemesser
Honored Contributor

Re: Rx8620 MP console hung

Hi,

You get absolutely nothing on the MP, right ? Or does only the console of your vpars not work, but you can enter something in f.e. the MP Main Menu ?
The "MP Reset" button makes a "soft reset" of the MP. Normaly this should fix a MP hang. But in some rare cases, the MP has to be powercycled (f.e. by powercycling the server or by unplugging and reinserting it).
In theory the MP can be unplugged online, but unfortunately, the same card hosts the Core I/O of cell 0 and if the partition that uses this Core I/O is up, then the system would crash with a HPMC.
The button on the top of the card is the OLAR button, but on rx8620 OLAR is not working. At least the partition with cell 0 must be down when unplugging the MP card.
Truxpin
Occasional Visitor

Re: Rx8620 MP console hung

Stefan, thanks for the useful reply.
Yes the MP does not work in any way.
The reset button did not help.
Yes I can access the running vPar via network.
Unfortunately I have one "stray" partition up using the SCSI adapters of the Core I/O module, but NOT any NIC (we used all the network cards for the production partitions).
This partition (I was setting as our ISEE partition but did not finish due to lack of NICs) is up and running but I do not kwow how to shut it down... is there a way to interact with the MON or this partition via the other parts?

Thanks again for your help (now I also know the meaning of the upper button!)
Stefan Stechemesser
Honored Contributor
Solution

Re: Rx8620 MP console hung

Hi,

the problem is that ALL vpars of the npar that has this Core I/O (cell 0) needs to be shut down, otherwise surely a MCA (machine check abort, all vpars of this npar down) will happen when you unplugg the Core I/O card.
Without the MP, it is dangeorous to shutdown your vpars, because if the MP is really defect, it may be that you cannot boot them anymore (because you have no console access and because EFI may hang during bootup if a bad MP is in the system).
In theory you could shutdown all vpars and finally shutdown the cells of this partition from another running npar on this server with the "frupower" command, but nobody knows if this works if the MP is brain dead.
Maybe it would be the best to open a case with HP support to have a replacement MP available if it is really bad.
Another option if after the powercycle of the Core I/O/MP (unplugging the power cables or unplugging the MP) the problem still is not fixed would be to swap the two Core I/O cards (the MP function on Core I/O 1 is unused).
Truxpin
Occasional Visitor

Re: Rx8620 MP console hung

Thankyou Stefan for having been sharp and clear about this issue.
We cannot afford an outage of this server and definitely we will have an MP replacement card before scheduling the downtime.
Just to make sure I've understood: can I swap the two MP in case the first is definitely dead?
It looks to me than in this config the Slave MP will not take over and we cannot force it to take over from the outside.

Thanks again for your support.
Truxpin
Occasional Visitor

Re: Rx8620 MP console hung

For the record:
we've closed all the running Vpars and powered the server down (actually we pulled the plugs!)... then we powered it back on and the Master MP came back to life - it kept the network config (for the sake of the many times I've pressed the "Reset MP" button!!!).
We had a service rep. present with a replacement Core I/O card ready and I strongly reccomend to do so if you ever happen to confront a similar problem: no MP no full power on and no Pars boot!
Thanks to all who contributed.