Re: Disable write caching

From: Scott Schnoll [MSFT] (scschnol_at_online.microsoft.com)
Date: 12/21/04

  • Next message: Peter Lykkegaard: "Q: Failover cluster and task schedule" 
    Date: Tue, 21 Dec 2004 10:44:46 -0800

    Yes, and they all call-out specifically using battery backed caches. 

    -- 
Scott Schnoll
This posting is provided "AS IS" with no warranties, and confers no
rights. Please do not send email directly to this alias. This alias is for 
newsgroup
purposes only.
"Rodney R. Fournier [MVP]" <rod@die.spam.die.nw-america.com> wrote in 
message news:uyY4Hr45EHA.3596@TK2MSFTNGP12.phx.gbl...
> Ok, so you know I am not crazy do this. Go to Google and use this - 
> site:hp.com write-caching Exchange.
>
> Notice all the CURRENT articles mention to set write-caching to 100%.
>
> Examples:
> http://h71019.www7.hp.com/aa_downloads/6/100/225/1/73833.pdf Page 28 100% 
> write
>
> http://h18002.www1.hp.com/solutions/messaging/medium_business_pop.html - 
> 642 controller with 100% write-caching.
>
> Lastly, HP Exchange 2000/2003 Best Practices - 
> http://h71019.www7.hp.com/aa_downloads/6/100/225/1/73843.pdf Page 24.
>
> Note: I am not really replying to Scott or Microsoft, but more to explain 
> my position.
>
> Cheers,
>
> Rod
>
> MVP - Windows Server - Clustering
> http://www.nw-america.com - Clustering
> http://msmvps.com/clustering - Blog
>
> "Scott Schnoll [MSFT]" <scschnol@online.microsoft.com> wrote in message 
> news:OE3BRh45EHA.1264@TK2MSFTNGP12.phx.gbl...
>> I'll weigh in.  :-)
>>
>> If you run an unprotected write-back cache against the log drive, logical 
>> corruption may occur. An unprotected write-back cache is a cache where 
>> data is lost if the electricity is interrupted (e.g., no battery backup). 
>> When a write operation to the log disk is complete, Extensible Storage 
>> Engine 98 (ESE 98) commits the transaction and makes the data durable on 
>> the disk. Any durable updates persist even if the computer stops 
>> responding immediately after the transaction is committed because the 
>> system's restart procedure completes any unfinished operations that the 
>> transaction requires. After the data is written to the log disk, ESE 98 
>> is free to flush the corresponding page to the database disk. If the data 
>> was never written to the log disk, the same problem occurs and the 
>> recovery process cannot restore the database to the state that it was in 
>> before the database was corrupted. Therefore, the database remains 
>> corrupted.
>>
>> The problems that may occur when you use write-back caching do not 
>> prevent you from using write-back caching. However, because the potential 
>> for problems does exist, you must make sure that the data in the cache is 
>> correctly protected with a battery backup, with error checking, with 
>> error correction, and with sound operational procedures.
>>
>> Please see the following references on this:
>>
>> http://support.microsoft.com/kb/288700
>> http://support.microsoft.com/kb/828068
>> http://support.microsoft.com/kb/329232
>>
>> Hope this helps.
>> -- 
>> Scott Schnoll
>> This posting is provided "AS IS" with no warranties, and confers no
>> rights. Please do not send email directly to this alias. This alias is 
>> for newsgroup
>> purposes only.
>>
>>
>> "Rodney R. Fournier [MVP]" <rod@die.spam.die.nw-america.com> wrote in 
>> message news:eoLcQa45EHA.3648@TK2MSFTNGP11.phx.gbl...
>>> Not according to HP at Tech Ed San Diego this year, he said to turn it 
>>> on now a days. Nothing about T-Log issues, which I have never seen. 
>>> Maybe someone from Microsoft will weigh in.
>>>
>>> Cheers,
>>>
>>> Rod
>>>
>>> MVP - Windows Server - Clustering
>>> http://www.nw-america.com - Clustering
>>> http://msmvps.com/clustering - Blog
>>>
>>> "Russ Kaufmann [MCT]" <russ@exchangemct.nospam.com> wrote in message 
>>> news:Oxx80S45EHA.1296@TK2MSFTNGP10.phx.gbl...
>>>> "TwistedPair" <twistedpair@mail.com> wrote in message 
>>>> news:ed%23G6L35EHA.3644@tk2msftngp13.phx.gbl...
>>>>> All,
>>>>> Would it be a true statement to say that write-caching could lead to 
>>>>> corruption on an Exchange 2003 server in a clustered environment?
>>>>
>>>> It was a known problem with 5.5 and also was a problem that I ran into 
>>>> several times with 2000 and 2003. Battery backups didn't make a 
>>>> difference in the cases that I saw.
>>>>
>>>> I recommend not using write back caching for the transaction logs. It 
>>>> seems to be the root of many evils. I have not seen the same problem 
>>>> with the DB, though, just the transaction logs.
>>>>
>>>
>>>
>>
>>
>
>
  • Next message: Peter Lykkegaard: "Q: Failover cluster and task schedule"

    Relevant Pages

    • Re: Slow copy to SBS 2003
      ... If you have a battery on the controller enable 'Write Back', ... cache is written to disk on shutdown). ... when copying TO the server. ...
      (microsoft.public.windows.server.sbs)
    • Re: Slow copy to SBS 2003
      ... If you have a battery on the controller enable 'Write Back', ... cache is written to disk on shutdown). ... most others) works at the sector level of the HDD's. ... when copying TO the server. ...
      (microsoft.public.windows.server.sbs)
    • Re: [dm-devel] Re: [PATCH] Implement barrier support for single device DM devices
      ... Using working barriers is important for normal users when you really ... write cache that will survive power outages... ... And second, "surprisingly", battery-backed RAID write caches tends to fail ... such a battery is enough to keep the data ...
      (Linux-Kernel)
    • SUMMARY: how to replace cache battery in HSG80
      ... one should bring the array down and flush the cache prior to replacing the ... battery slots for each controller. ... ordered and received a replacement battery, but I'm not exactly sure what ...
      (Tru64-UNIX-Managers)
    • [TEST] LTTng relay-alloc cache tests
      ... to keep the buffer pages as write-back than to set them to ... LTTng trace buffer memory cache behavior comparison ... [timing 1, cache cold] ... sys 0m48.943s ...
      (Linux-Kernel)