Does Teamsite creates its own inodes dynamically

Hi All,

Version of TS :- 7.1 on linux.

We had a problem few days back when the default mount was suddenly getting unmounted. The reason autonomy gave was that we were crossing inodes limit (max 11 million). But when i checked for inodes below is the result.

Filesystem Inodes IUsed IFree IUse% Mounted on
none 151915556 41125300 110790256 28% /iwmnt/default

So, I have no idea where autonomy came up with the answer that the TS file system is running out of inodes.

When I pressed for documentation/explanation on this below is the answer

*****************

"/iwmnt issue is related to TeamSite internal inodes, which are not exposed through any file system monitoring tools. The inode numbers shown below are forwarded from the backing store file system.
Compare the df output for /iwmnt/default with the df output of the backing store file system to see what I mean.

TeamSite dynamically creates inodes as needed, and keeps a record of the most recently used ~11 million inodes. There are no external tools that can be used to determine when a particular TeamSite inode will become the least recently used inode and be discarded."
***************

As per my knowledge Inode is an attribute of a file/folder in unix irrespective of the file size.i.e if i create a file of say 5gb it will have only one inode tagged to it. If i delete the file the inode attached to it will be free.

Can any one explain what are " Teasmite Internal Inodes" ? 2. How inodes are created dynamically in Teamsite ?

This is the 1st time I am hearing that in Teamsite Inodes are created dynamically. I have searched several forums and found that inodes are not created dynamically but we can increase the inodes count. Here is an example of such blog http://blog.colovirt.com/2009/04/10/linux-unix-nas-file-systems-inodes-part-1-checking-availability-and-high-level-overview/

Any explanation on this will help me in learning.....

Cheers
Cool_rash

Find more posts tagged with

Comments

ISCBorisB

Inode is an attribute of a file/folder...if i create a file of say 5gb it will have only one inode tagged to it

iNode is not a File attribute. You can have any number of iNodes consumed by a single File.

File's iNodes collection is a Linked List (Meta)Data structure that contains certain information about File
Blocks. "ls -i..." shows you a root Node. iNode number within iNode Table *maps* to the physical location on
the disk so that kernel can access node's content.

Single iNode represents fixed number of Blocks. Actual number is normally set when you format new disk. When
this set is full, another iNode is linked to the list allowing the File to grow.

All that aside, please remember that iwmount is a virtual drive. iwserver actually "fools" OS to believe that
iwmount exists by intercepting and processing corresponding system calls. How exactly they do it is not disclosed.
If they say they use "internal iNodes" (whatever that means), they probably do. What if anything this control
structure has in common with actual Linux iNodes, how (if?) it uses them and all that are all proprietary
implementation details.

cool_rash

I agree to that but a df -i on the virtual mount is telling that 28% of inodes are only consumed.
So, its difficult to "believe" that during operation on the mount the inodes count is increasing more than 200 %.....which is leading for the mount to unmount...Autonomy gave us a Workaround to restart the Teamsite on a daily basis.....

This is for the first time I am facing this issue and autonomy isnt able 2 explain this scenario...
I have worked with all the previous version of TS starting from 5.6 in earlier projects having unix platform but never faced issue with inodes. Is it something related to 7.1 ????

ISCBorisB

I'm not sure if my reading of the cited Support reply is any better than yours but still...

They basically make two points:
1. "issue is related to [...] internal inodes, which are not exposed through any file system monitoring tools" and
2. "compare the df output for /iwmnt/default with the df output of the backing store..."

To me it sounds like your df's 28% (or whatever) usage is irrelevant. You are not looking at "internal iNodes" and in fact
have no tools to do so. As to the Support recommendation - have you tried it? What does df show for the real OS mount
that keeps TS backing store?

Rick Poulin

TeamSite on Linux uses UVFS (User-space Virtual File System), which is an implementation of a file system that runs on top of the host's file system and it therefore uses its own inodes internally. These seem to be represented externally by the d0-0 folders in some fashion. As iwov support said, I'm not aware of any diagnostic tools that would allow you to inspect the state of the UVFS instance, but it certainly has very little to do with the host file system's allocated inodes. Maybe a count of the files/folders under the content store folder would be some indication of how many inodes you've used up, but I'd expect that the closest you can get to a diagnostic is the iwfsck CLT.

Regardless of all that, if you have in fact maxed out the content store's inodes, it may be time to start archiving your oldest editions.

cool_rash

Autonomy is checking on this. I guess we will have a solution and will definitely post here....

Thanks for all the help.

Cheers
RashmiRanjan