Page Deallocation Table (PDT)

[jamshed]

What I have to do?

The Page Deallocation Table (PDT) is 100% full.

PDT Entries Used: 50
PDT Entries Free: 0
PDT Total Size: 50

An excessive number of memory pages have been deallocated due to
excessive correctable single bit errors being detected. Although the
errors are being corrected, this condition indicates a potential
problem.

[Gary L. Paveza, Jr.]

Most likely you have a failing memory DIMM. You will need to contact your support vendor (HP most likely).

[jamshed]

If I lost my memory DIMM, can I lose my whole system? To tell the truth I have not any DIMM for replacing. Not now.

Can I work in this way? Until I find another module?

[Gary L. Paveza, Jr.]

You haven't lost the DIMM yet. What has happened is that the system has marked at least 50 pages (4k) as unusable and removed them from use. I say at least 50 because that's the size of the PDT table. There could be more than 50. The DIMM itself is still in use. However, it is indictitive of a failing DIMM - so it is possible for the entire DIMM to fail at some point in the near future. You can certainly continue to run. If you are under maintenance, then you shouldn't have any problem getting your vendor to replace. If you aren't, then you may want to look at purchasing a new DIMM soon.

[Sundar_7]

There is a option to clear PDT entries.

1) Reboot the system

2) Interrupt the autoboot

3) go to configuration menu

I believe you have the option to clear PDT entries from here. I cannot recollect the exact command though.

[Bill Hassell]

If all the errors are for one memory DIMM, shutdown the machine and remove that DIMM and it's pair (DIMMs must always be installed or removed in pairs). You memory is going bad and depending on where the an uncorrectable error occurs, it will likely crash the system. Correctable errors (lots of them) usually indicate that a hard, uncorrectable error is close at hand.

[Stefan Stechemesser]

Hi,

the PDT entries are genertated by the online diagnostics if single bit errors happen. If an excessive amount of errors happen on one dimm, then it points to a hardware problem on this Dimm. But: we should keep in mind that these are only single bit errors. Single bit errors are corrected without delay and the only impact on the performance is that CPU time is used for logging the error via the dm_memory monitor and the memlogd process. The same diagnostic processes are responsible for writing PDT entries. This is only a preventive action to decrease the chance of a double bit error.
Although the chance of a double bit error is higher than before, it is still very low, because this would mean that two bits flip in one byte at the same time which is not very probable in a Dimm > 1 MB.
Let's say we have two independend single bit errors in 1 MB. The chance that after the first single bit error, the next error happens on the same byte is ~ 1000000:1 and you need a lot of single bit errors to produce such an error.
Unfortunately it may also be that the errors are not independent and in this case it may indeed be that the server will crash in the future with a double bit error (and if a double bit error occurs during the selftests, and the PDT is full, then the Dimm will be deallocated).

If you want to save your money and take a risk then ignore the errors. Maybe your system will run for the next 100 years without a crash ;-)
If a double bit errors occur, then you must replace the dimm (depends on server type if one or two dimms have to be replaced).

good luck

Stefan

[Thayanidhi]

You shouldn't clear PDT unless you replace the DIMMS. If you clear without replacing, System will reuse the area which are already known as bad. Again the PDT will get full.

Most recommended now is to replace the bad memeory module.

TT