Raw Device

[former_member1338072]

Hello,

I am planning to migrate an existing maxdb to a new server hardware. I thinking about using raw devices for the data volumes. My question is, has using raw devices real advantages on linux over "normal" file system in conjunction with the parameter USE_OPEN_DIRECT ?

Thanks

Tiberius

[lbreddemann · Accepted solution]

Hi Tiberius,

the advantage should be rather neglectable. When using filesystems you will always have a bit overhead due to filesystemmanagement and you've to be careful about which filesystem you choose.

On the other hand - what do you want to do with MaxDB volumes in the filesystem? There's no backup-per-filecopy as it is with Oracle databases.

Therefore rawdevices are giving you everything you really need for MaxDB databases. And as these are less complex (no os/filesystem overhead) they are less error prone and easier to manage than filesystems (ok, sure you've to actually know how to manage them).

regards,

Lars

[markus_doehr2 · Accepted solution]

We are running our MaxDBs on RAW devices for a long time (8+ years) and it worked very well. We had cases where after a crash the automatic fsck "repaired" a volume - only apparently - because it was hosed up after that.

Markus

[former_member1338072 · Accepted solution]

thank You for the answers. I am not new to linux and MaxDB, but I never used raw devices until now. Where can I find a tutorial how to setup maxdb with raw devices ? Is 1 raw device=1 data volume ? Are there special parameters to set/change ? Or should I give one big partition to maxdb ..?

Thanks

Tiberius

[lbreddemann · Accepted solution]

Hi Tiberius,

yes, one raw device will be one MaxDB Volume.

Usually you would use the raw command on Linux to configure the raw-device access.

After this is done (so that the raw devices are usuable via /dev/rawN) you can simply use the DBMGUI/DB Studio to add the volumes to the database.

regards,

Lars

[markus_doehr2 · Accepted solution]

We use "partitions" - for a 2 TB database we create 4 x 500 GB LUNs and use (c)fdisk to create 14 partitions. Then those are mapped vi /etc/raw (on SuSE) from block to character devices.

Markus

[former_member1338072 · Accepted solution]

ok, I will try my luck.

regards

Tiberius

[Former Member · Accepted solution]

Hi, Markus,

There is a way to you share a step by step on how to do that partition? i'm interested because all my installation i do using ext3, and like to follow some best patterns and i think you have one.

best regards.

Clóvis

[nelis · Accepted solution]

Have you tried [Google|http://tinyurl.com/yb9tx3e] ?

Nelis

[Former Member · Accepted solution]

Hi, Nelis,

You have that knowledge? can you share a step by step installation on linux from scratch, from partition, to maxdb configuration? for example how you configure a machine to be used only for maxdb, that have dual quad core processors, and 6 SAS hardisk, (3 RAID 1 will be mounted) and 24GB of ram?

all my searchs goes to SAP Note, but i dont have a password to read SAP note, as i'm from community, is too frustating that some good knowledge are only inside SAP notes, why SAP dont share that with community too? or put that info in documentation?

best regards, and please, next time reply with a really usefull answer.

Clóvis

[lbreddemann · Accepted solution]

> all my searchs goes to SAP Note, but i dont have a password to read SAP note, as i'm from community, is too frustating that some good knowledge are only inside SAP notes, why SAP dont share that with community too? or put that info in documentation?

Well, maybe it's because raw-disk-configuration is not a topic of SAP software?

Once the raw-device is properly installed, the MaxDB configuration part is just the same as for filesystem data volumes.

regards,

Lars