Conference smurf::buildhelp

85.0. "ODE *Down-Time* Friday 6-10am" by SMURF::FILTER (Automatic Posting Software - mail to flume::puck) Mon Jun 07 1993 16:48

Date Of Receipt: 	27-MAY-1993 09:15:19.24
From: 	MINSRV::"[email protected]" "Joshua M. Friedman ULTRIX SDE"
To: 	[email protected]
CC: 	[email protected]
Subj: 	ODE *Down-Time* Friday 6-10am

OSF/1 Developers both in zk and in other sites (except doc groups),

This Friday morning, 5/28, from 6 - 10am, we have scheduled maintenance
"down-time" for all the zk3 production build hosts serving:
    - /usr/sde tools,
    - build trees,
    - submit pools, and
    - rcs pools.

This will affect all release streams, both for individual workstations
and production machines.

This scheduled activity is one of a series of actions being taken to
help improve users' build performance, reduce network load, and improve
production availability of backing trees.

To summarize the changes, mounts of project build trees will change from
/usr/sde/disks/<mnt>/<project> to /share/<hostname>/build/<mnt>/<project>.

> Prior to the down-time (Thursday afternoon/night) all zk3 users should:

    * Ensure that no user sandboxes directly point to the physical
	mount points in /usr/sde/disks.  All sandboxes created
	using the normal "mksb -back <project>" command are ok.
	(You can "resb <project>" to properly retarget a sandbox.)

    * cd out of /usr/sde trees on production machines

    * unmount all backing and submit trees (odemount -u -all)
	(you can first use "odemount -list" to capture what's mounted)

    * umount /usr/sde

    * change your /etc/fstab to get /usr/sde from your subnet's ode server
	(see /usr/sde/disks/ode.servers)

    * For improved isolation change any PATH definition references
	from `/usr/sde/toolsbin.sh` to /usr/sde/tools/`machine`_`uname`/bin

    * Note that nightly build schedules for the release streams will be
	adjusted to accomodate for this down-time

> During the down-time, Friday morning:

    * /usr/sde will be unavailable from zk3 servers

    * backing trees for all releases will be unavailable from zk3 servers

    * No bco/bci/bsubmit commands will function from ANY SITE
	(for the OSF/1 pools "base os" and "x11" pools)

> After the down-time (after announcement is mailed on Friday) users should:

    * remount /usr/sde from your local subnet's ode server

    * use odemount to remount the required build trees

    * Note that automount will no longer serve /usr/sde areas in zk3

    * Note that you can choose to reboot your systems after the work is
	completed, if you don't want to bother to unmount and remount
	everything.  Your /etc/fstab will have to be updated if you
	have mounted any /usr/sde/disks directly, and should reflect
	the correct /usr/sde server for your subnet.

FYI, in the near future, we will be working with the RMT and project
teams to schedule routine maintenance down-time for the build-net
machines, for example, perhaps 2 hours every other Friday.  This will
be a consistent time to allow for planning.

If you have any high-risk issues with this, please contact your manager.

For a detailed plan of the work being done and the reasons it is
required, see the file (on any zk3 production machine):

    /usr/specs/release_bld/plans/share_mounts.plan

Thank you very much...

		    --Josh Friedman & DEC OSF/1 Release Engineering Team

Date Of Receipt: 	27-MAY-1993 15:51:40.20
From: 	MINSRV::"[email protected]" "Joshua M. Friedman ULTRIX SDE"
To: 	[email protected]
CC: 	[email protected]
Subj: 	ODE *Down-Time* reminder

This is a reminder for users to prepare for the down-time tomorrow 
morning of ODE servers.  Before you leave tonight zk3 users should:

    * cd out of /usr/sde trees on production machines

    * unmount all backing and submit trees 

    * umount /usr/sde

    * change your /etc/fstab to get /usr/sde from your subnet's ode server
        (see list attached)

The mail which was sent describing the details is on production hosts in:

    /usr/specs/release_bld/plans/share_mounts.user-mail


FYI, the nightly build schedule for tonight has been revised as follows:
    - maintenance os: already running
    - maintenance x: will start tomorrow after work is completed
    - sterling os: will start 3:50 pm (i.e. now)
    - sterling x: no new submits so no build tonight
    - gold os: will start 5:30 pm
    - hardware: no new submits so no build tonight

Thank you very much...
			    --josh & release engineering


------------------------------------------------------------
For your reference, here are the current /usr/sde servers on each
subnet.  (This information will be maintained in
/usr/sde/disks/ode.servers and in the netinfo database.)

    oskits:     red-yellow-net (IP address 16.140.0.*)
    xor:        yellow-net (IP address 16.140.16.*)
    lastout:    green-net (IP address 16.140.32.*)
    firstout:   green-net (IP address 16.140.32.*)
    bigbld:     green-white-net (IP address 16.140.48.*)
    lastin:     red-net (IP address 16.140.64.*)
    *flambe:     blue-yellow-net (IP address 16.140.96.*)
    buffer:     orange-net (IP address 16.140.112.*)
    *ruddy:      blue-net (IP address 16.140.128.*)

*available only after the downtime tomorrow

Date Of Receipt: 	28-MAY-1993 10:48:42.09
From: 	MINSRV::"[email protected]" "Joshua M. Friedman ULTRIX SDE"
To: 	[email protected]
CC: 	[email protected]
Subj: 	ODE Down-Time Completed

The upgrade of the ode mount environment in zk has been completed, and
you can now re-mount /usr/sde from your server (see attached list),
and odemount projects.  

The pools have been unlocked.

Please report any problems to [email protected] .

There is a caveat:  due to having reached a 256 filesystem limit, for
the production machines:

	krisis, wasted, flume, and abyss

the only ode trees which are mounted are:

	agosmaint, agxmaint, agosminor, agxminor, goldos, agoshw,
	agosmaint.nightly, agxmaint.nightly, agosminor.nightly,
	agxminor.nightly, goldos.nightly, agoshw.nightly,
	alpha.bl012, and x11.bl012

Please mount /usr/sde from your local subnet ("arp hostname" will
show your ip address; match it to the ones in this list):

    oskits:	red-yellow-net (IP address 16.140.0.*)
    xor:	yellow-net (IP address 16.140.16.*)
    lastout:	green-net (IP address 16.140.32.*)
    firstout:	green-net (IP address 16.140.32.*)
    bigbld:	green-white-net (IP address 16.140.48.*)
    lastin:	red-net (IP address 16.140.64.*)
    flambe:     blue-yellow-net (IP address 16.140.96.*)
    buffer:	orange-net (IP address 16.140.112.*)
    ruddy:      blue-net (IP address 16.140.128.*)

thank you very much...
			    -Josh Friedman / Release Engineering / UES Admin