[LU-14751] Prevent users from setting dir and file stripping Created: 09/Jun/21 Updated: 01/Jul/21 |
|
| Status: | Open |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | None |
| Fix Version/s: | None |
| Type: | New Feature | Priority: | Critical |
| Reporter: | Mahmoud Hanafi | Assignee: | Li Xi |
| Resolution: | Unresolved | Votes: | 0 |
| Labels: | None | ||
| Rank (Obsolete): | 9223372036854775807 |
| Description |
|
A filesystem with multiple tier of storage may require specific PFL. We may not want users to change these default. There need to be an option to prevent changing default stripping. |
| Comments |
| Comment by Peter Jones [ 10/Jun/21 ] |
|
Li Xi What do you think about this suggestion? Thanks Peter |
| Comment by Li Xi [ 10/Jun/21 ] |
|
Mahmoud, do you mind to discribe what would happen if the PFL setting is different from the required one in your environment/usage? In the multiple tier storage environment, misconfiguration of PFL could indeed lead to bad performance, waste of high-speed storage media, or/and exhaustion of small-capcity tier. Any problems that are more serious problem than these? Are we thinking of preventing some thing that is really bad for global users from happening? I am not sure that is the case according to the current information. I do agree that the default PFL striping should be carefully configured. And if the file system has the multiple tiers, even more care should be spent on the default setting. After configuring the default PFL properly, users should be able to inherit the default configuration and benefit from it. Unless something that is really bad will happen, I feel that we should provide the flexibilty of changing PFL settings to the users. It is unlikely that a user who has little knowledge of PFL and the tier structure would change the setting by themselves. And if they choose to change something that will only hurt the performance of their own applications a little bit, they should have the freedom to learn the lesson. And who knows, maybe they will figure out a even better PFL configuration for their own usage. So, what do you think? |
| Comment by Mahmoud Hanafi [ 14/Jun/21 ] |
|
Our biggest concerns are exhaustion of small capacity tier/OSTs. So we would want to prevent this type of issue. Usually a very small set users that would gain benefit from setting stripe different than default. |
| Comment by Mahmoud Hanafi [ 23/Jun/21 ] |
|
In our experience correct usage/string doesn’t lead to ost exhaustion unless the filesystem is very full. OST quota will not prevent users or application from in correctly striping an SSD tier. |
| Comment by Mahmoud Hanafi [ 01/Jul/21 ] |
|
I like to get an update. |