University of Cambridge Computer Laboratory - Archive SMB service migration – Maintenance details

Text of the email announcement:

This message is meant for researchers who store data on the archive server, archive.cl.cam.ac.uk.  Professional services staff can ignore this announcement.  Researchers who aren't sure whether they use archive can see below for some guidance on how to tell.

The archive server, archive.cl.cam.ac.uk, is being replaced.  The server will be unavailable on one or two occasions (first one: 4th-5th July) whilst this work takes place.  You may then need to change how you access the server.

How do I know if I am using archive?

Most people are not using archive.  However if you contacted us asking for backup space or more than a few hundred gigabytes of storage at any point during the last several years, we might have provided you with storage on archive.  We would generally have discussed this with you when creating the storage, though in some cases that would have been many years ago.

The archive server is not related to filer or bigdisc.

Generally any access to archive will be through the hostname “archive.cl.cam.ac.uk” or simply “archive”, or paths including the component "archive" or "gfxdisp".  (The server name "berilia" is also involved; if you are using anything referring to "berilia" please contact sys-admin@cl.cam.ac.uk as this will stop working.)

Archive provides two separate services: NFS (Linux/UNIX style) shares and SMB (Windows-style) shares.  (SMB may also be referred to as CIFS.  SMB can be accessed from Linux clients too, though mostly it's intended for Windows.)  At the moment, both services use the same server name, archive.cl.cam.ac.uk.  We will be migrating both services to a new server, but on different days.  However, all users of archive will be impacted to some extent on that day, and almost all users of archive will need to change how they use the service.

You are using archive’s SMB (Windows-style) service, a.k.a. CIFS, if any of the following applies:

• You are using path that look similar to any of the following:

  • \\archive\something or \\archive.cl.cam.ac.uk\something (Windows UNC paths)

  • smb://archive.cl.cam.ac.uk/something or smb://archive/something (Linux/Mac file browser SMB paths)

  • //archive.cl.cam.ac.uk/something or //archive/something (UNC paths as used by some Linux clients such as “mount -t smbfs” or “mount -t cifs”)

• You are using “corelab_datasets” (this is a special case; see below)

• You are using “scy27_traces” (this is a special case; see below)

If any of these apply, see "Impact on SMB users" below; expect a long period of disruption on 4th July, perhaps extending into 5th July, and to have to change how you access archive.

You are using archive’s NFS (Linux/UNIX-style) service if any of the following applies:

• You are manually mounting storage from archive.cl.cam.ac.uk:/export/something or archive:/export/something, or have one of those paths in /etc/fstab

• You are using paths under /auto/archive or /net/archive or /anfs/gfxdisp or /auto/anfs/gfxdisp

If any of these apply, see "Impact on NFS users" below; expect a short disruption on 4th July, then another email about a longer disruption a month or two later.

Impact on SMB users

The SMB service will move first, on 4th July.  It will be unavailable for several hours on that day, potentially with disruption extending to the next day due to the large amount of work needed to reinstate this service on a new server.

Immediately after the work on 4th July, the archive SMB service will move to archive-smb.cl.cam.ac.uk.  You will need to update any path that you use that currently refers to the SMB service on archive.cl.cam.ac.uk.

Access permissions on SMB shares will be reset!  For each SMB share, we have identified the current users and will grant those users read and write access to the entire share.  Every other member of the department will be given read access (in order to minimise disruption, as we believe there is no confidential data on archive).  If you have confidential data on archive that must not be readable by other users, please contact sys-admin@cl.cam.ac.uk immediately.

Any custom permissions that may have been set on individual files and folders will be lost.  Unfortunately it is not feasible to transfer permissions from the old server to the new one, as we are switching software platform (from Spectra Verde to TrueNAS) and the two platforms store Windows-style ACLs in different, incompatible ways.

Whilst the permissions are being rebuilt on 4th-5th July – which is a manual process – you may find that you are unable to access your storage.  We will try to restore read access first; write access may take longer.

If you are using corelab_datasets: This is a special case; it is a SMB (Windows) share, but several members of your group access it from managed Linux systems via a temporary mechanism.  We plan to make this share available via both SMB and NFS, to better support access from both operating systems.  We will reconfigure each managed Linux machine on which we’ve set up access to this dataset to use NFS, once this volume is available via NFS.  Until then, you will be unable to access corelab_datasets from Linux.

If you are using scy27_traces: As discussed with the users (RT ticket #135996), this is currently a NFS share but is being migrated along with the SMB shares so that we can enable SMB access in future.  So this will experience the same disruption as SMB shares, described above, including the reset of permissions.  On 4th July it will also start to require a Kerberos ticket for NFS access.

Impact on NFS users

Firstly, on 4th July the NFS service will be unavailable for a short while (approximately 30-60 minutes) as the whole archive server will need to be shut down in order to move the disks holding SMB data to another server.  Preparatory work has already been completed to try to expedite this process; however there is a chance of unexpected problems with bringing the NFS service back online on the old server (due to complications with the old server's obsolete software platform).  We will prioritise bringing the NFS service back up as quickly as possible on 4th July.

There will be a future more-disruptive outage to archive’s NFS service in a month or two when we move that service to the new hardware; the timeline for this is not yet decided.

But if you are accessing an archive NFS filesystem from a Lab-managed Linux/UNIX system, you can prepare now for the migration by ensuring that you are not using a path starting /net/archive.  If you are using /net/archive/export/SHARE please switch to /auto/archive/SHARE as soon as convenient.

If you are accessing an archive NFS filesystem from a non-Lab-managed computer by manually mounting archive[.cl.cam.ac.uk]:/export/SHARE, we suggest that you email sys-admin@cl.cam.ac.uk now to ask that we help you to set up the Lab automounter on your computer.  If you don’t, you’ll have to change to a different path when we move the NFS service.

At the moment, NFS shares on archive only support “sec=sys” (IP-address-based access); they do not use any form of strong authentication.  After the upgrade, we will be in a position to optionally authenticate NFS using Kerberos, as we do on filer.  Contact sys-admin@cl.cam.ac.uk if you are interested in switching your archive NFS share to Kerberos.

Updates during the work

Updates on progress during the work (as well as a copy of this announcement) will be posted at https://cl.instatus.com/clxj8zjas110017bkln1e6boqwu .  You can subscribe on that page if you would like to receive updates by email.

Please contact sys-admin@cl.cam.ac.uk if you have any questions.  Thanks.