MegaCli Using write through instead of write back

Question

My DB server is a dell r710 with megacli. Recently was getting poor performance from the database and stumbled across that the current cache policy is WT but the default cache policy is WB.

# ./MegaCli64 -LDInfo -Lall -aAll | grep 'Cache Policy:'
Default Cache Policy: WriteBack, ReadAdaptive, Direct, No Write Cache if Bad BBU
Current Cache Policy: WriteThrough, ReadAdaptive, Direct, No Write Cache if Bad BBU

Looking some though some of the other articles on here, is only occurs when the battery is faulty. Looking at the BBU it looks like its good.

# ./MegaCli64 -AdpBbuCmd -GetBbuStatus -aALL | grep isSOHGood
isSOHGood: Yes

Am I missing something? How can I change this to be WriteBack so I can get the performance back.

Thanks!

Solution 1

You can switch back the policy back to WB with

megacli -LDSetProp WB -LAll -aAll

After that you can check the current state with

megacli -LDInfo -LAll -aAll

Solution 2

Use the force luke

megacli -LDSetProp -ForcedWB -Immediate -Lall -aAll

I wouldn't set Forced WriteBack under heavy load, but your main problem right now seems to be a faulty BBU so I would double check it. Normally the BBU should be replaced every 1-5 years and regulary checked.

If

megacli -AdpBbuCmd -GetBbuStatus -aALL

really doesn't say anything about learning required or a faulty battery than maybe a firmware upgrade is required.
To check the version:

megacli -AdpAllInfo -aAll

To update the firmware:

MegaCli -adpfwflash -f firmware.rom -a0

Be carefull and good luck with your raid.

Comments

[deagh]

Some controllers change the policy while battery learning cycle and switch back after that. Maybe your controller missed that switch back.

[screwuphead]

I have seen that as well but looking at the logs and at the BBU Fireware status the learn cycle is not active. "Learn Cycle Active: No"

[screwuphead]

Already tried enabling writeback and trying to force it with the -Immediate. No change in the current cache policy.

[sjas]

do you have dell? might be that the command create jobs in the idrac job queue, which you can run then