My favorites | Sign in
Project Home Downloads Wiki Issues Source
READ-ONLY: This project has been archived. For more information see this post.
Search
for
  Advanced search   Search tips   Subscriptions
Issue 48: Testing With OHSM patch
1 person starred this issue and may be notified of changes. Back to list
 
Project Member Reported by imreckless@gmail.com, Jan 30, 2009
We started with testing.

Yesterday we patched the kernel.
Then there were few changes in new_blocks function
we removed reservation window.
We re-patched the kernel with new changes.

There were errors in ohsm.c
We rectified that.
We had to clean the ohsm.c code, it was very messy.
We then created a patch for the same.

Kernel was patched and compiled successfully.
We inserted the ohsm_admin module.
It gave undefined symbol sam 

In balloc.c bg_range was defined but in admin sam was being used
So we changed bg_range in balloc.c to sam
We have done the changes in balloc.c and ohsm.c

We thought that we need to check all the errors, 
then create the new patch accordingly.

The kernel is compiled successfully now with changes.
We have created 8 4Gb loop devices
We have created 4 tiers, each with 2 devices.

sam table contained all 0 values.
nr_targets was showing 0
we rectified it, dev/vg/lvol0 was not opening.

Now admin is inserted successfully.

We created allocation and tier info xml file
Then we compiled userspace program

We provided all the parameters to CLI
it was working fine

At this point things were fine.

We tried creating files.
We created directories and files.

When we tried creating files in vim it gave filesystem full error.
Unexpected behaviour, some time it creates file sometime it doesnt.

We are rectifying this error.
We are now applying reservation window code again.

Update you soon.

Jan 30, 2009
Project Member #1 imreckless@gmail.com
We tried creating files with dd,
system crashed when we created 400 Mb file.
rebooting

Jan 30, 2009
Project Member #2 sandeepksinha
I need Updates after every 30 mins on this issue.

This work is at the highest priority.
See if the code for opening device is working fine ?
Any clues why it was not working?

The device should open....
Did we require to do a mknod for the LV also ? Not sure. But AFAIK, we don't.


Jan 30, 2009
Project Member #3 imreckless@gmail.com
We tried to create small files, it worked.
System is crashing with large files.

Hdd corrupt, we are reconfiguring everything again.
No clues yet.

We have moved to other system for testing.
Still its not allowing to create physical volume on hdd.
Reformatting the hdd.

Jan 30, 2009
Project Member #4 imreckless@gmail.com
System configured, starting the test
4 devices 4G each /dev/sdb1 /dev/sdb5 /dev/sdb6 /dev/sdb7
tier1 /dev/sdb1 
tier2 /dev/sdb5 /dev/sdb7
tier3 /dev/sdb6

Total block groups 128, 16G Logical volume.

policy 
file type
.c ->  tier1
.cc -> tier2
.ccc -> tier3
.cccc-> tier1

test started, showing nr_targets 4
Sucess.

error in sam table.
Problem with script, it was working with loop device.
I think SKS and Rishi has a better solution for this script.
We have changed the script to work for scsi disks.

Getting segfaults. Problem with sam table.
Rectifying it.

Jan 30, 2009
Project Member #5 checkout...@gmail.com
The Problem with sam table has been rectified.

Rishi is currently working with us.


Jan 30, 2009
Project Member #6 imreckless@gmail.com
Now it is allocating 1G file but for the files that do not fall in allocation policy.

Few more changes are in inode.c, ext2_iget and ext2_update inode.
Ohsm fields home_tid, dest_tid, fiot and fat must be synced to disk inode
and retrieved from disk inode. This is done now.

Printing information, bg_start and bg_end fields, and tid in ohsm_ext2_new_blocks

Expected problem in sam table.
start_bg = 127 end_bg = 0
Thats why its giving no space error.
Rectifying it soon.





Jan 30, 2009
Project Member #7 imreckless@gmail.com
sam table problem has been rectified.
The sizes in balloc was MAX_DEV 255 and MAX_DEV_PER_TIER 5
and in admin it was 10 and 10 respectively.

Now allocation is working.
System is still crashing.
Rectifying this problem currently.
Status: Inprogress
Jan 30, 2009
Project Member #8 imreckless@gmail.com
Allocation successful.
Created 4Gb file, qualified policy and allocated blocks in specified tier.

Moving on to relocation.

Jan 30, 2009
Project Member #9 imreckless@gmail.com
Well we relocated 4G file, its was relocated successfully.
We relocated small files also it relocated successfully without diff.

Moving on to rigorous testing with large number of files. 
Jan 30, 2009
Project Member #10 checkout...@gmail.com

Testing of ohsm patch is successful.

Now a working Ohsm Patch is ready.
Status: In-progress
Jan 30, 2009
Project Member #11 imreckless@gmail.com
Kernel compiled with OHSM patch successfully.
Rigorous testing successful. 
Around 300 files relocated successfully without diff.
Allocation works, relocation works.

Attaching a log file, just ls and diff output.
Will generate new logs, the log was messy with all the debugs enabled.

Will proceed testing with different disks.
Collecting scsi disks of different types. :P

And i will make sure that it crashes.

We are almost OHSM enabled.
Congrats Team !!


test log
14.2 KB   View   Download
Status: Fixed
Jan 31, 2009
#12 postri...@gmail.com
Good Job Everybody............. we still need SKS to run it. 

making status Review Request
Status: Review-Req
Labels: -Type-Defect Type-Task OpSys-Linux Design-Confirm-Code Usability
Jan 31, 2009
Project Member #13 checkout...@gmail.com
Thanks Greg for pointing out.

we have successfully relocated an running mp3 file. 

There was a slight pause in playback during relocation.
Owner: checkout.vineet
Jan 31, 2009
Project Member #14 imreckless@gmail.com
Test using different disks, and check the access time.
Owner: imreckless
Jan 31, 2009
Project Member #15 sandeepksinha
So, Guys we would require to test our changes with SQL. We need a concrete test plan
for the same.

Also, remember that mp3 was just in read only mode and hence we don't expect much
complications.
Performing rigorous read/write will be the only way to test this.
I think somebody needs to do a good amount of work to build a test suite.

RKS can you take this work up ?
I think we already have a issue open for the same.


Feb 26, 2009
Project Member #16 imreckless@gmail.com
Patch is tested successfully, i m closing this issue
Status: Done

Powered by Google Project Hosting