Closed Bug 690611 Opened 9 years ago Closed 9 years ago

Run rev4 10.6 slaves concurrently with rev3 10.6 and have graph server results for both


(Release Engineering :: General, defect, major)

Not set


(Not tracked)



(Reporter: jhford, Assigned: jhford)




(2 files)

I'd like to have rev4 10.6 slaves running at the same time as rev3 slaves.  I'd rather not have them report as the same platform as they are completely different hardware.

I was thinking of doing:

1) rename "MacOSX 10.6.2" to "MacOSX 10.6.2 (rev3)"
update os_list set name = "MacOSX 10.6.2 (rev3)" where id = 17;
2) create "MacOSX 10.6"
insert into os_list values (NULL, "MacOSX 10.6");
-- fix, because the auto-incrementer is wrong
3) have talos-r4 machines report to this new platform
insert into machines values (NULL,<ID_FROM_NEW_OS_ID>,0,"2.4","talos-r4-snow-001",1,unix_timestamp());
insert into machines values (NULL,<ID_FROM_NEW_OS_ID>,0,"2.4","talos-r4-snow-183",1,unix_timestamp());

I have some concerns with this:

1) Will this break links?  My understanding is that links have specific data point ids, not strings to the platform.

2) Is this a valid thing to do? is there a better way we could do this?
This patch is the changes to sql/data.sql needed to recreate our db
This script is designed to fix existing production database
Also, another question, would I need to change the graphserver post at all to have the rev3 machines point to the updated platform -- or would that happen as a result of the relation between and machines.os_id?
Blocks: 690236
Severity: normal → major
Comment on attachment 563606 [details]
sql script to rename old mac 10.6

Do either of you guys have thoughts on this?
Attachment #563606 - Flags: feedback?(rhelmer)
Attachment #563606 - Flags: feedback?(catlee)
Comment on attachment 563606 [details]
sql script to rename old mac 10.6

Looks fine, as long as you are sure about the ids :)

Where are all the values for the inserts coming from? If they are already in the db could do this with insert and a subselect but nothing wrong with doing it explicitly like this for such a small set.

Also a INSERT can insert multiple rows, if you want to make it less verbose:

 INSERT INTO tbl_name (a,b,c) VALUES(1,2,3),(4,5,6),(7,8,9);

Not important though.
Attachment #563606 - Flags: feedback?(rhelmer) → feedback+
Its been a while since I have written a subselect and I don't remember the syntax :(.  I am pretty sure about the IDs, but I don't think a quick |select * from os_list where name = "MacOSX 10.6.2"| check wouldn't hurt.

The inserts that I have are really just the patch I have above filtered through sed to remove the + at the beginning of the line and the diff headers.

Since it looks like everything is in order, I'll ask for r? on this patch and script now.  Thanks for the quick feedback!
Attachment #563605 - Flags: review?(rhelmer)
Comment on attachment 563606 [details]
sql script to rename old mac 10.6

I will hold off on deploying this until the r4 machines are ready for production
Attachment #563606 - Flags: review?(rhelmer)
Attachment #563606 - Flags: feedback?(catlee)
Attachment #563606 - Flags: feedback?
Attachment #563606 - Flags: review?(rhelmer) → review+
Attachment #563605 - Flags: review?(rhelmer) → review+
Depends on: 692113
This went into production today.  I have verified that rev3 results are using the new OS name.
Closed: 9 years ago
Resolution: --- → FIXED
Product: → Release Engineering
Attachment #563606 - Flags: feedback?
You need to log in before you can comment on or make changes to this bug.