Subscribe

Details on NVRAM Journal Replay after an Outage

[ Edited ]

Dear all,

 

I have been studying the basics of ONTAP and WAFL recently. Discussions on the Interent are very old and even official TR-3002 was moved away. I read several materials saying that

 

1. NVRAM is used to store JOURNAL but not DATA so it can be small but just enough

2. Data are wriiten to disk at each Consistent Point (CP), and between CPs data only go to DRAM with journal going to NVRAM

3. In case of power outage, journals are kept in NVRAM waiting to be replayed as the power come back

 

I wonder what WAFL is to do if power outage result in disappearance of data in DRAM; with only the journals, can WAFL simply finish what was not done yet?

Any reply would be appreciated. Thanks in advance.

Re: End to End vs WAFL drilldown

Nachin -

 

You seem to have the basics correct.

 

ONTAP would read the pending writes from NVRAM on boot and commit the CP to disk...

 


I hope this response has been helpful to you.

 

At your service,

 

Eugene E. Kashpureff, Sr.
Independent NetApp Consultant http://www.linkedin.com/in/eugenekashpureff
Senior NetApp Instructor, FastLane US http://www.fastlaneus.com/
(P.S. I appreciate 'kudos' on any helpful posts.)

 

Re: End to End vs WAFL drilldown

Dear Sir, Thanks for your instant reply. I am not sure why I have a wrong title. I am asking for details of NVRAM journal replay anyway. The question is as ONTAP boots to finish the journal, if previously there are write I/Os and the data cached in DRAM has not been flushed to disk before the outage, how would WAFL do as it sees the journal? Where can it find the lost data, or I misunderstand something here? Thank you.

Re: End to End vs WAFL drilldown

 

Nachin -

 

NVRAM is backed up by a battery.

 

ONTAP would read those journaled writes from NVRAM into RAM and then complete the writes to disk.

 


I hope this response has been helpful to you.

 

At your service,

 

Eugene E. Kashpureff, Sr.
Independent NetApp Consultant http://www.linkedin.com/in/eugenekashpureff
Senior NetApp Instructor, FastLane US http://www.fastlaneus.com/
(P.S. I appreciate 'kudos' on any helpful posts.)

 

Re: End to End vs WAFL drilldown

 

Nachin -

 

Not sure why they took down TR-3002, but TR-3001 is still out there:

 

http://www.netapp.com/us/media/tr-3001.pdf

 


I hope this response has been helpful to you.

 

At your service,

 

Eugene E. Kashpureff, Sr.
Independent NetApp Consultant http://www.linkedin.com/in/eugenekashpureff
Senior NetApp Instructor, FastLane US http://www.fastlaneus.com/
(P.S. I appreciate 'kudos' on any helpful posts.)

 

Re: End to End vs WAFL drilldown

Dear Sir,

 

I apolpgize for being a little annoying, but this is what confused me. ONTAP does read journaled writes, but how does it complete the writes to disk?

If a few blocks are arrived at DRAM and write requests are journaled, and at the mean time power outage occurs so blocks in DRAM are lost. After ONTAP boots and wants to relpay the journaled writes, where can it find the lost blocks? Am I missing anything important?

Thank you.

Re: End to End vs WAFL drilldown

Writes that arrive at the controller are not "aknowledged" back to the host until they are also stored in NVRAM.

 

Just to be clear, the NVram does store the write data as well as the journal - In case of a power failure, the NVRAM is used to complete the writes that were aknowledged before the power went down.