Closed
Bug 598259
Opened 14 years ago
Closed 14 years ago
Set up access and routing to machine in Phoenix
Categories
(mozilla.org Graveyard :: Server Operations, task)
Tracking
(Not tracked)
RESOLVED
FIXED
People
(Reporter: nthomas, Assigned: aravind)
References
Details
RelEng needs to test pushing AUS snippets to a machine in the Phoenix data center, so please set up ssh access on a machine where we can dump up to a gigabyte of files. You could use the stage-ffxbld key in bug 455567.
We'd like to be able to reach the PHX machine from dm-ausstage01.m.o, staging-stage.build.m.o, and machines in the build-vpn (eg mv-moz2-linux-ix-slave01.build.m.o). Once AUS in PHX is set up we can arrange access for production systems and revoke this testing access.
This is for a joint IT & RelEng Q3 goal, which we'd like to work on Wednesday while RelEng are all in MV.
Reporter | ||
Comment 1•14 years ago
|
||
It'll be a dense tree of of small files, ~ 88K directories and nearly 140K of files for a base copy of the existing snippets, plus 25K dirs and 42K of files for each copy of new snippets we copy in. I know we've bumped the inode limit on dm-ausstage01:/opt so please double check that will be set large enough on the partition you're allocating, allowing for at least 1 base copy and 5 set of new snippets (210k dirs plus 350 files).
Reporter | ||
Comment 2•14 years ago
|
||
That should be s/of files/files/ in comment #1 - ie 140K and 42K individual files.
Comment 3•14 years ago
|
||
Easiest way to do this is probably through a VM in Phoenix that has an NFS mount to the filer there.
Ravi - can you get this going? Copying Phong to help spin up a VM.
Assignee: server-ops → ravi
Severity: major → critical
Comment 4•14 years ago
|
||
dp-ausstage01 (to keep consistent naming) / 10.8.74.30 is online.
Needs to be driven by puppet and needs stage-ffxbld account/keys.
Reporter | ||
Comment 5•14 years ago
|
||
ping, we'd like to use this machine today.
Comment 6•14 years ago
|
||
Flow has been opened between the two hosts.
[root@dm-ausstage01 ~]# nc -vz 10.8.74.30 22
Connection to 10.8.74.30 22 port [tcp/ssh] succeeded!
Status: NEW → RESOLVED
Closed: 14 years ago
Resolution: --- → FIXED
Reporter | ||
Comment 7•14 years ago
|
||
(In reply to comment #0)
> We'd like to be able to reach the PHX machine from dm-ausstage01.m.o,
> staging-stage.build.m.o, and machines in the build-vpn (eg
> mv-moz2-linux-ix-slave01.build.m.o). Once AUS in PHX is set up we can arrange
> access for production systems and revoke this testing access.
Could you please allow traffic between the staging slaves listed above and the dp-ausstage. Or you can set the access list like dm-ausstage01, in that all the build machines can access it (regardless of location).
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Reporter | ||
Comment 8•14 years ago
|
||
dp-ausstage01 is also refusing the ssh key for stage-ffxbld. I've checked I'm using the private key that matches the public one in bug 455567.
Updated•14 years ago
|
Assignee: ravi → justdave
Comment 9•14 years ago
|
||
ravi had one part, punting over to justdave for the rest.
Comment 10•14 years ago
|
||
Not familiar with how this is set up. What user are you logging in with for the ssh? I see an ffxbld user, is it that one?
Reporter | ||
Comment 11•14 years ago
|
||
We'll be doing our testing with our staging machines so it makes sense to use our staging accounts right now, or at least the staging key. So .... ffxbld with the key in the bug 455567, thanks!
Comment 12•14 years ago
|
||
ok, account created, key set up, give it a shot.
Reporter | ||
Comment 13•14 years ago
|
||
I can log in by ssh from dm-ausstage01, but can't reach 10.8.74.30 from staging machines, ie comment 7. We can start with dm-ausstage in the meantime though.
Comment 14•14 years ago
|
||
Please enumerate all the sources and the destinations by FQDN or IP and the ports you will be connecting to. This will help me write the config only twice.
Reporter | ||
Comment 15•14 years ago
|
||
I am asking for
destination: 10.8.74.30:22
sources: 10.2.71.82 staging-stage.build.mozilla.org
10.2.71.105 moz2-linux-slave03.build.mozilla.org
10.2.71.18 moz2-linux-slave04.build.mozilla.org
10.250.49.199 mv-moz2-linux-ix-slave01.build.mozilla.org
This will be for testing purposes. If the 10.250 is a PITA then you drop that and we'll deal.
Once we know what we'll be doing with production then we will want to change the list of sources.
Comment 16•14 years ago
|
||
I added the 4 additional hosts and am waiting for verification.
Reporter | ||
Comment 17•14 years ago
|
||
(In reply to comment #16)
They all work except 10.250.49.199 mv-moz2-linux-ix-slave01.build.mozilla.org, but that'll be enough. Thanks.
Reporter | ||
Comment 18•14 years ago
|
||
(In reply to comment #3)
> Easiest way to do this is probably through a VM in Phoenix that has an NFS
> mount to the filer there.
AIUI we use the netapp in MPT because it is HA, and mount via NFS to the webheads. I would prefer to test against this configuration rather than use the much faster storage on /. On dm-ausstage01 it takes us more than 30 minutes to create a nightly backup, and at least 5 mins to unpack it (it's still running as I write); vs ~ 30 seconds to unpack on dp-ausstage01. That's going to make a big difference to testing rsyncs from MPT to PHX.
Updated•14 years ago
|
Assignee: justdave → ravi
Comment 19•14 years ago
|
||
(In reply to comment #18)
> (In reply to comment #3)
> > Easiest way to do this is probably through a VM in Phoenix that has an NFS
> > mount to the filer there.
>
> AIUI we use the netapp in MPT because it is HA, and mount via NFS to the
> webheads. I would prefer to test against this configuration rather than use the
> much faster storage on /. On dm-ausstage01 it takes us more than 30 minutes to
> create a nightly backup, and at least 5 mins to unpack it (it's still running
> as I write); vs ~ 30 seconds to unpack on dp-ausstage01. That's going to make a
> big difference to testing rsyncs from MPT to PHX.
Many things:
1. dp-ausstage01 is missing an NFS mount off the NetApp in Phoenix
2. dp-ausstage01 & the Phoenix NetApp are going to be faster than anything in SJC.
3. Ironically, dp-ausstage01:/ is on the NetApp in Phoenix. It's just fast.
4. For POC, you can keep writing to / until someone gets the NFS mount for you.
Assignee: ravi → aravind
Comment 20•14 years ago
|
||
Nick/Rob - do you guys have what you need?
Reporter | ||
Comment 21•14 years ago
|
||
mrz says we're OK with the current disk setup and we have access for the staging machines. That will be fine for testing.
Status: REOPENED → RESOLVED
Closed: 14 years ago → 14 years ago
Resolution: --- → FIXED
Comment 22•14 years ago
|
||
Still need the NAS setup - Aravind's point on that (but this doesn't block POC testing).
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Assignee | ||
Comment 23•14 years ago
|
||
The mount is ready. Its commented out in fstab.
10.8.74.240:/vol/pio_aus2/ /opt nfs noatime,rsize=8192,wsize=8192,intr 0 0
Status: REOPENED → RESOLVED
Closed: 14 years ago → 14 years ago
Resolution: --- → FIXED
Reporter | ||
Comment 24•14 years ago
|
||
(In reply to comment #12)
> ok, account created, key set up, give it a shot.
FTR, the username is ffxbld, and the key is stage-ffxbld.
Updated•10 years ago
|
Product: mozilla.org → mozilla.org Graveyard
You need to log in
before you can comment on or make changes to this bug.
Description
•