mounting SSD devices of EC2 r3.8xlarge instances

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

mounting SSD devices of EC2 r3.8xlarge instances

Andras Barjak
Hi,
I have noticed that upon launching a cluster consisting of r3.8xlarge high-memory instances the standard /mnt /mnt2 /mnt3 /mnt4 temporary directories get created and set up for temp usage, however they will point to the root 8Gb filesystem.
The 2x320GB SSD-s are not mounted and also they are not even formatted.

This problem might affect other EC2 instances as well, I suppose.
I am using 0.9.1, is this something that has been corrected in the 1.0.0 spark-ec2 script?

regards,
András Barják
Reply | Threaded
Open this post in threaded view
|

Re: mounting SSD devices of EC2 r3.8xlarge instances

Matei Zaharia
Administrator
Those instance types are not yet supported by the scripts, but https://issues.apache.org/jira/browse/SPARK-1790 is tracking this issue and it will soon be fixed in both branch-0.9 and 1.0. The problem is that those drives are not formatted on r3 machines, whereas they are on the other instance types, so the script assumed that they’d be formatted. You can manually go and do this to set them up:

mkfs.ext4 /dev/sdb
mkfs.ext4 /dev/sdc
mount -o noatime /dev/sdb /mnt
mount -o noatime /dev/sdc /mnt2

Matei

On Jun 3, 2014, at 10:05 AM, Andras Barjak <[hidden email]> wrote:

> Hi,
> I have noticed that upon launching a cluster consisting of r3.8xlarge high-memory instances the standard /mnt /mnt2 /mnt3 /mnt4 temporary directories get created and set up for temp usage, however they will point to the root 8Gb filesystem.
> The 2x320GB SSD-s are not mounted and also they are not even formatted.
>
> This problem might affect other EC2 instances as well, I suppose.
> I am using 0.9.1, is this something that has been corrected in the 1.0.0 spark-ec2 script?
>
> regards,
> András Barják

Reply | Threaded
Open this post in threaded view
|

Re: mounting SSD devices of EC2 r3.8xlarge instances

darkjh
For SSDs in r3, maybe it's better to mount with `discard` option since it supports TRIM:

What I did for r3.large:

  echo '/dev/xvdb /mnt ext4 defaults,noatime,nodiratime,discard 0 0' >> /etc/fstab
  mkfs.ext4 /dev/xvdb 
  mount /dev/xvdb



2014-06-03 19:15 GMT+02:00 Matei Zaharia <[hidden email]>:
Those instance types are not yet supported by the scripts, but https://issues.apache.org/jira/browse/SPARK-1790 is tracking this issue and it will soon be fixed in both branch-0.9 and 1.0. The problem is that those drives are not formatted on r3 machines, whereas they are on the other instance types, so the script assumed that they’d be formatted. You can manually go and do this to set them up:

mkfs.ext4 /dev/sdb
mkfs.ext4 /dev/sdc
mount -o noatime /dev/sdb /mnt
mount -o noatime /dev/sdc /mnt2

Matei

On Jun 3, 2014, at 10:05 AM, Andras Barjak <[hidden email]> wrote:

> Hi,
> I have noticed that upon launching a cluster consisting of r3.8xlarge high-memory instances the standard /mnt /mnt2 /mnt3 /mnt4 temporary directories get created and set up for temp usage, however they will point to the root 8Gb filesystem.
> The 2x320GB SSD-s are not mounted and also they are not even formatted.
>
> This problem might affect other EC2 instances as well, I suppose.
> I am using 0.9.1, is this something that has been corrected in the 1.0.0 spark-ec2 script?
>
> regards,
> András Barják




--
JU Han

Data Engineer @ Botify.com

+33 0619608888