PFL self-extending file layout (LU-10070)

[LU-12365] Create test plan for SEL Created: 31/May/19  Updated: 25/Sep/19  Resolved: 25/Sep/19

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Technical task Priority: Blocker
Reporter: Cory Spitz Assignee: Vitaly Fertman
Resolution: Fixed Votes: 0
Labels: None

Attachments: PDF File LUSTRE-SelfExtendingLayoutTestPlan-090719-1129-72.pdf     PDF File LUSTRE-SelfExtendingLayoutTestPlan-v0.1.pdf     File local.sh     File test_script.sh     File test_script_dom.sh     File test_script_plain.sh    
Rank (Obsolete): 9223372036854775807

 Description   

A test plan for SEL needs to be created in support of merging the code to master.



 Comments   
Comment by Cory Spitz [ 31/May/19 ]

Thank you vitaly_fertman. Can you please update the Feature Interaction section to include FLR and Overstriping?

Comment by Cory Spitz [ 31/May/19 ]

pfarrell has written the attached scripts, which demonstrate the functionality. He's provided a guide:

There are three test scripts and a test config attached. The config is “local.sh”.
These demo scripts are written for a single node running the sanity config. (They assume the MDS/MGS are on the same node.)
On CentOS7, setup would look like this:
Replace:
/usr/lib64/lustre/tests/cfg/local.sh with the local.sh attached here
Run:
/usr/lib64/lustre/tests/llmount.sh
To set up the local file system.
cd /mnt/lustre
Then exec the test scripts, in any order. They print out what they’re doing and ask for the user to hit enter to continue.
The main test_script.sh shows off the basic SEL extension functionality using pools. (This is a difference from the demo script.)
The test_script_dom_pools.sh uses DOM & pools to show off skipping a tier if it is already full.
The test_script_plain.sh script shows off the normal file case – A simple SEL layout going to EOF. This layout can switch to new OSTs if the ones it’s using get low on space. This is a potential “default” layout for a whole file system, with or without tiering.

Comment by James Nunez (Inactive) [ 03/Jun/19 ]

Thank you for the Self Extending Layout test plan. I have a few questions/comments on the attached, version 0.1, of the SEL test plan:
PFL is “Progressive File Layout”, not “Partial File Layout” as stated in the Introduction.

In the Introduction, the link for more information to SLE, https://connect.us.cray.com/confluence/display/LUSTRE/HLD%3A+Spillover+Space+-+Self-Extending+Layouts , does not work. Would you please put this information somewhere that we can access?

Under Regression Tests, you specify that some Lustre test suites will be run with different SEL layouts set on the ROOT_DIR. Would you please include a link to those test results in the test plan or in this ticket?

Similarly, would you please add the results of your small IO, large IO and MD performance testing to the test plan or to the ticket?

Under Feature interaction/Snapshot, should a SEL test be added to sanity-lsnapshot.sh to cover the SLE/snapshot interaction?

Is the functionality of the attached scripts from Patrick incorporated into some of the automated test scripts? If not, are there reasons why they aren’t incorporated?

Comment by Patrick Farrell (Inactive) [ 04/Jun/19 ]

Note the 'demo' scripts attached have not been updated to use -z instead of -x.  So they need that update at least, and it would be good for Cray to re-run those and make sure they still work with current versions, as various things have changed.

Comment by Cory Spitz [ 04/Jun/19 ]

jamesanunez, sorry about the internal link. The HLD is attached to the main ticket at https://jira.whamcloud.com/secure/attachment/30498/LUSTRE-98540817-120718-1412-90.pdf.

Comment by Vitaly Fertman [ 07/Jun/19 ]

Thank you for the Self Extending Layout test plan. I have a few questions/comments on the attached, version 0.1, of the SEL test plan:
PFL is “Progressive File Layout”, not “Partial File Layout” as stated in the Introduction.

heh, right

Under Regression Tests, you specify that some Lustre test suites will be run with different SEL layouts set on the ROOT_DIR. Would you please include a link to those test results in the test plan or in this ticket?

you will not have rights to open that link however we will definitely provide some testing summary report once it will be done. not as a part of the TP doc itself though.

Similarly, would you please add the results of your small IO, large IO and MD performance testing to the test plan or to the ticket?

to the ticket, once done, some summary report will be posted.

Under Feature interaction/Snapshot, should a SEL test be added to sanity-lsnapshot.sh to cover the SLE/snapshot interaction?

sure, it should be tested at the same level as PFL. so where can I find the PFL/snapshot interaction tests - I do not see them in this script? do I get it right this feature is already officially maintained ?

Is the functionality of the attached scripts from Patrick incorporated into some of the automated test scripts? If not, are there reasons why they aren’t incorporated?

no idea, I see them for the 1st time. if they are a kind of functional feature tests, this is already covered by new sanity tests.

Comment by Cory Spitz [ 09/Jul/19 ]

For reference -
Old test plan: LUSTRE-SelfExtendingLayoutTestPlan-v0.1.pdf
Updated test plan: LUSTRE-SelfExtendingLayoutTestPlan-090719-1129-72.pdf

Generated at Sat Feb 10 02:51:57 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.