PDT full

[G.Carslow]

Hi,

We are running a L1000 with dual proc and 2Gig RAM. The PDT is filling with entries indicating slots 0a/0b memory cards have excessive bit errors. This was reported to HP and they ( after investigation ) replaced the cards in 0a/0b. The problem has occured again showing much the same errors. After clearing the PDT and reboot the PDT is empty for a couple of days.
Any input would be appreciated as this is a critical machine.
Thanks
Greg

[T G Manikandan]

Looks like it is working fine.
The PDT must have been cleared during the replacement.Closely monitor them for another fortnight.

[G.Carslow]

Hi,

Thanks but the problem has occured again since the replacement of the memory modules. I have concluded from this that a problem still exists.

Thanks
Greg

[Stefan Stechemesser · Accepted solution]

Hi Greg,

This error is not critical, because single bit errors neither produce crashes or corruptions nor is there a performance impact, because the correction algorithm will be performed for every byte read from memory. The PDT entries are made by the EMS software. A crash would only occur if two bits in one byte would be wrong. This is not very probable in a 2 Gig RAM system unless there is a real hardware problem (like two IC pins are broken).
However the cause of this type of problem is (decreasing chance) either a bad Dimm, a bad seated dimm, a bad dimm connector on the system board or a bad memory controller (Astro chip, also on the system board).
I would recomend to call HP again and let them continue the repair. It is normal NOT to swap a system board due to uncritical memory problems in the first try and you should plan at least a 2h downtime if HP and you decide to replace the system board.

best regards

Stefan

[hector hernandez]

Hi Greg.

The problem has reappeared in the same pair of DIMM's ? Do not change it again, just yet. Try first to swap the pair of DIMM's to another slot (Ex. 1a/1b) and clear the PDT. If the problems moves with the DIMM's then change the DIMM's, it probably was a DOA. If the problem remains in the same slot then the most probable cause is the ASTRO chip. Although it is not common, and in very corner cases the problem also can be a software application, but it is very unlikely.

Regards.

Hector

[G.Carslow]

Hi,

Thanks for replies. At this point it is causing the application ( HPOV NNM ) to crash. I think logging with HP again is probably the best idea.

Thank you
Greg