Hi Nirmal,
Please see my comments inline.
On 02/15/12 10:12 AM, Nirmal Agarwal wrote:
Hi Karen
I tested the scenario's you suggested. Below are my findings and
suggested solution(s) :
On 2/15/2012 12:35 AM, Karen Tung wrote:
Hi Nirmal,
Dave's question below prompted me to think of a couple of
questions on the proposed bug fix:
1) Since size is now optional, I can have more than 1 zvol
where I don't specify the size. Based on your fix, each of those
zvols that does not have a size will get 90% of the zpool size. That
obviously
won't work. How would this situation be handled?
As per the current implementation, it creates 2nd zvol as 90 % of
remaining pool size left after creation of 1st zvol (which is
effectively 9 % of zpool).
We can handle this situation in following manner :
--> allow only 1 zvol with undefined size.
--> allow undefined size zvol for "swap" use only.
--> create the swap in last to allocate 90 % of available space in pool.
Among the 3 choices above, I like the choice to allow undefined zvol
size for "swap" use only
the best. I don't like the first choice, because it seems kinda odd to
allow only 1 zvol with
undefined size in general. I don't like the 3rd choice because we still
can have more than 1 zvols
with undefined size, and anything besides the first undefined size zvol
is pretty much useless.
2) A related question, if I specify a zpool that have both ZFS filesystems
and a zvol where I didn't specify a size, is it necessary for us to
warn the
user that the ZFS filesystem that will be created will only be 10%
size of the pool.
I will introduce a "warn" level debug message whenever we create zvol
with 90 % of the pool size.
A "warning" level debug message during execution is certainly useful.
In addition,
I think we can improve the user experience as follows:
1) Since you are changing the target.dtd, put some comment in the section
where the zvol size is optional, and discuss what will happen if the user
did not specify a size. That way, people won't be surprised when they
do the
actual install.
2) If we provide any sample AI manifests for users to use as starting
points,
also put some comment there talking about the behavior.
Thanks,
--Karen
Please let me know your comments on the approach.
Thanks
Nirmal
Thanks,
--Karen
On 02/14/12 08:28, Nirmal Agarwal wrote:
Hi Dave
I did test the full installation and it works.
Let me know if you want to run any other tests.
Thanks
Nirmal
On 02/14/12 03:22, Dave Miner wrote:
I haven't reviewed the code here, but can you elaborate on the
testing?
Does AI create a correctly usable swap volume, and will Solaris boot
with that swap volume once you've done so? It's not entirely clear
to me
that this will work.
Dave
On 02/13/12 00:12, Nirmal Agarwal wrote:
Hi all
Could I please get a code review for the following CR:
7070697 size element for zvols should be optional
Webrev :
https://cr.opensolaris.org/action/browse/caiman/nirmal27/7070697/webrev/
slim_test result
----------------
/net/indiana-build.us.oracle.com//export/home/na210770/ai/7070697/slim_test
Manual Tests :
-- created a custom image and tested the fix with the manifest not
describing the size of zvol.
Thanks
Nirmal
_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
_______________________________________________
caiman-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
