XFS+noatime for Zookeeper

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

XFS+noatime for Zookeeper

Joshua Dotson (josdotso)
Hello.

Is there any reason using an XFS filesystem mounted with the noatime option would be a problem for Zookeeper data (/var/lib/zookeeper).  From my reading, this XFS+noatime is a good match for Kafka, so I’m just wondering if this will hurt/enhance Zookeeper in any way.

I’m using the container mentioned here:

https://github.com/kubernetes/contrib/tree/master/statefulsets/zookeeper

Thanks,
Joshua
Reply | Threaded
Open this post in threaded view
|

Re: XFS+noatime for Zookeeper

Raúl Gutiérrez Segalés
Hi,

On Thu, Mar 29, 2018 at 5:12 PM, Joshua Dotson (josdotso) <
[hidden email]> wrote:

> Hello.
>
> Is there any reason using an XFS filesystem mounted with the noatime
> option would be a problem for Zookeeper data (/var/lib/zookeeper).  From my
> reading, this XFS+noatime is a good match for Kafka, so I’m just wondering
> if this will hurt/enhance Zookeeper in any way.
>
> I’m using the container mentioned here:
>
> https://github.com/kubernetes/contrib/tree/master/statefulsets/zookeeper


fwiw, I've been using like that -- in production -- for years:

/dev/md0 on /zkdata xfs
(rw,noatime,nobarrier,discard,inode64,logbsize=256k)


-rgs
Reply | Threaded
Open this post in threaded view
|

Re: XFS+noatime for Zookeeper

Sargun Dhillon
On Thu, Mar 29, 2018 at 5:21 PM, Raúl Gutiérrez Segalés
<[hidden email]> wrote:

> Hi,
>
> On Thu, Mar 29, 2018 at 5:12 PM, Joshua Dotson (josdotso) <
> [hidden email]> wrote:
>
>> Hello.
>>
>> Is there any reason using an XFS filesystem mounted with the noatime
>> option would be a problem for Zookeeper data (/var/lib/zookeeper).  From my
>> reading, this XFS+noatime is a good match for Kafka, so I’m just wondering
>> if this will hurt/enhance Zookeeper in any way.
>>
>> I’m using the container mentioned here:
>>
>> https://github.com/kubernetes/contrib/tree/master/statefulsets/zookeeper
>
>
> fwiw, I've been using like that -- in production -- for years:
>
> /dev/md0 on /zkdata xfs
> (rw,noatime,nobarrier,discard,inode64,logbsize=256k)
>
>
> -rgs
Likewise. But a suggestion. Use relatime. It'll make debugging in the
future without much of a performance hit. If you're using a modern
enough kernel, the metadata overhead is negligible.