Limit the Number of files in Cache1/retr folder

[GURU]

Hi All,

can any one help me with the below concern please

Is there any way to limit the number of files (or links) in Cache1/retr folders, so that it can moved to some other Cache(2-4)/retr folder(Creating retr folder manually). Because we have recently migrated our CMOD server to different data center with separate App server & DB server after which we are facing some space issue as we are using NAS storage.

Below are our server details:
OD Version: 8.5.0.6
DB: Oracle 11g
OS version: RHEL




 

[Justin Derrick]

There isn't a way to specifically limit the number of files in a cache.  You can keep forcing expiration to use a smaller and smaller percentage of the available space -- but leaving empty space in a cache filesystem is wasteful.

If you add a new cache filesystem, newly loaded files will go there first, as CMOD tries to distribute loaded files evenly across all cache filesystems.

For the record, you can only have ONE "retr" directory in your cache filesystems, and it must be the first cache filesystem.  Adding more cache filesystems won't reduce the size (or number) of links in the retr directory!

Could you clarify the problem you're experiencing?  Are you running out of storage space (gigabytes) or are you approaching the maximum number of files that can be stored in a filesystem (inodes)?

-JD.

[SV]

Guru,

Can you explain what was the issue you want to reduce the number of links? I am Curious, I heard this type of issue before some thing to do with inodes; but there is no clear documentation on how many files can be created. IF you explain your issue might be useful.

[Alessandro Perucchi]

Some explanation would be nice.

But as already said... no way to limit the number of files/link in retr. If you have a problem with it, check the limitation of your filesystem in your current setup, and change the filesystem to have one which handle the things better with less limitations.

[Steve Bechtolt]

To add to Justin's comment about the 'retr' links in the first defined cache filesystem; CMOD reserves 5% of the first cache filesystem for this.
So when the first cache filesystem hits 95% utilized, no new storage objects will be written to that filesystem, as long as it is at 95% or above utilized.

[Justin Derrick]

To add to Justin's comment about the 'retr' links in the first defined cache filesystem; CMOD reserves 5% of the first cache filesystem for this.
So when the first cache filesystem hits 95% utilized, no new storage objects will be written to that filesystem, as long as it is at 95% or above utilized.

Interesting... thanks for the tip! 

-JD.

[dbenson100]

We are seeing a similar issue

$ df -g /arsca*
Filesystem         GB blocks      Free %Used    Iused %Iused Mounted on
/dev/arscache01lv     14.00      1.52   90% 23906568    99% /arscache01
/dev/fslv02          1499.00   1279.57   15% 23995306     8% /arscache02

I know we have  big discrepancy between the size of cache01 and 02, but we thought the system would limit the files added to 01 before it filled up.  Last week the 01 filesystem completely filled and our AIX support team added (4) more GB.  I also need to have them expand the inodes still.

Is there anything we can do to prevent this filesystem from filling up again?

[pankaj.puranik]

To add to Justin's comment about the 'retr' links in the first defined cache filesystem; CMOD reserves 5% of the first cache filesystem for this.
So when the first cache filesystem hits 95% utilized, no new storage objects will be written to that filesystem, as long as it is at 95% or above utilized.

Interesting... thanks for the tip! 

-JD.

Here's a tech note on this topic.
http://www-01.ibm.com/support/docview.wss?uid=swg21409251

Cheers
Pankaj.

[pankaj.puranik]

We are seeing a similar issue

$ df -g /arsca*
Filesystem         GB blocks      Free %Used    Iused %Iused Mounted on
/dev/arscache01lv     14.00      1.52   90% 23906568    99% /arscache01
/dev/fslv02          1499.00   1279.57   15% 23995306     8% /arscache02

I know we have  big discrepancy between the size of cache01 and 02, but we thought the system would limit the files added to 01 before it filled up.  Last week the 01 filesystem completely filled and our AIX support team added (4) more GB.  I also need to have them expand the inodes still.

Is there anything we can do to prevent this filesystem from filling up again?

Did you check if there is a permission issue?
All the cache directories should be owned by the same ID that runs ARSSOCKD.

Cheers
Pankaj.