Trouble at t' mill

[Wacky]

At around 1500 UT today, Monday 26th January, the NFSNET server here at
MSR Cambridge UK lost its disk which holds all the NFSNET data. Neither
the operating system nor the BIOS can see that disk any longer. This is
not good news 8-(

Richard Wackerbarth and I have worked to restore service. The data
disk here is still unavailable so we put a skeleton server into
operation on the system disk. There is just enough information in place
to allow us to redirect all the clients which were taking tasks from the
Cantab and the Fenland pools to their counterpart pools which are hosted
on Richard's server in Austin, Texas. At the time of typing it's about
20 minutes after setting up the redirections and several dozen clients
have already transferred.


Now that the emergency work has been completed, I will concentrate on
getting my server back into operation. I will try very hard to regain
access to the disk and its data because there is approximately 15 hours
worth of sieving from about 125 clients on it which is not stored
anywhere else. All data which arrived before 0000UT today has already
been copied to Austin for safekeeping.

Apologies for the drop in service. We were relatively fortunate in that
it occurred during the working day in both Austin and Cambridge, that we
noticed in good time and that we had an emergency response procedure
prepared and ready to be activated.


More news when we have it.


Paul & Richard

[junky]

that's bad :(
I've a question about pool:
i've remark sometime is Fenland, sometime is Texas. what that mean ?
it is just the place where its getting the info from ?
thanks.

[Jeff Gilchrist]

The pool is the server you are connecting to in order to get assignments. There are several pools setup (in the UK and Texas). If there is a problem with one server (like above) people will be redirected to another so they can continue to process work. It is mainly used to spread out the load from all the users.

[dsouza123]

Maybe a safety feature of storing the last 24 hours/last dozen results.txt files on the PC so if the work is lost on the server it could be resent.

The client could tell the server the date/time of it's stored files if they are old ( hadn't done work in a while) or they are securely on a server ( if I understand correctly, the work gets backed up to another server roughly daily) they are deleted.

On my PC Athlon 1200 1.5 hours = 24kb results file so 24 hours worth would be about 384kb.

[junky]

RAID5 isn't there for that purpose ?

[xilman]

The server at MSR Cambridge is now serving again. Almost all
NFSNET clients are being moved back to the Cantab and Fenland
pools. Several dozen have already transferred.

I replaced the failed disk this morning and restored data from
a backup.

The old disk did spin up again when I plugged it into another
machine and it is now approximately 80% of the way through
copying all its data on to another machine. So, apart from a
couple of hours yesterday between the disk dying and everyone
being moved over to a Texas server, we haven't lost any data.

It will take a little time before the data on the old disk makes
its way into the warehouse and the stats appear on the web pages.
With luck, they may appear tomorrow. Otherwise, they should be
there on Thursday.

Again, apologies for the interruption in our service.


Paul

[junky]

hi paul,
im now getting that output:

17:54:40 NFSNET Client - V 1.0
17:54:40 Initializing...
17:54:40 Initialized.
17:54:40 Requesting assignment...
17:54:43 Received assignment (  IDLE  0-0)...

why its idle? still getting the info from the backup??

thanks.