Start a Conversation

Unsolved

6 Posts

2279

April 2nd, 2020 22:00

isilon inode will full ,but filesystem still have a lot of space, why? thanks !

isilon inode will full ,but filesystem still have a lot of space, why?   who can help me ,thank you very much!

i have 5 nodes, NL400. onefs 8.0.1.2. 

[root datanode04 tdh]# df -i /tdh
Filesystem                           Inodes        IUsed                   IFree          IUse%      Mounted on
:/tdh/datenode04       245618442240  201995782016 43622660224   83%          /tdh
[root datanode04 tdh]# df -h /tdh
Filesystem                      Size  Used     Avail     Use%      Mounted on
:/tdh/datenode04             30T  9.7T        21T      33%             /tdh
 
[root datanode04 tdh]# for i in `ls -1A | grep -v "\.\./" | grep -v "\./"`; do echo "`find $i | sort -u | wc -l` $i"; done | sort -rn | head -10
123614 hadoop
3614 gpdata
696 nmon
1 data04

Moderator

 • 

7.1K Posts

April 3rd, 2020 17:00

Hello haoyang.wang,

I would call into support so that we can look into your system to see what all is going on.

Please let us know if you have any other questions.

3 Apprentice

 • 

592 Posts

April 3rd, 2020 19:00

inodes are dynamically allocated in Isilon, so you should never worry about inodes.

 

https://www.dellemc.com/en-us/collaterals/unauth/white-papers/products/storage/h16857-wp-onefs-best-practices.pdf    page 1

OneFS dynamically allocates new inodes from free file system blocks

April 3rd, 2020 20:00

but,Continue to write data. When the inode is full,  cannot continue to write data.

April 3rd, 2020 20:00

 

WechatIMG34.png

I get these alerts all the time

WechatIMG36.pngWechatIMG38.png

Continue to write data. When the inode is full,  cannot continue to write data.

3 Apprentice

 • 

592 Posts

April 7th, 2020 20:00

please open an SR then.

April 8th, 2020 00:00

But my isilon is out of warranty.

1.2K Posts

April 8th, 2020 07:00

The reporting of inode statistics to a NFS client via the df command has changed over various releases of OneFS, and also depends on quota settings. In your original example, the displayed inodes counts are huge, and obviously have nothing to do with the actual inodes in the files system.

As already noted, capacity for inodes is shared with file data capacity, so it is not possible to run out of inode space before file space is exhausted.  

You might set up a small test folder with quota and export it, and try out filling it up with smaller or larger files to see how the df inode reports behave in your environment. In the end, this might help you to savely ignore your application's alerts; or possibly to adjust the alerting settings -- AFAIK ElasticSearch is somewhat configurable.

hth

-- Peter

 

No Events found!

Top