[evla-sw-discuss] [ngas-alerts] Upgrade EVLA Lustre clients Tomorrow

[K. Scott Rowe]

Everything is back up.  Please let me know if you see any problems.

I had to power cycle cbe-node-16 and cbe-node-21 but otherwise the CBE cluster
rebooted correctly and is now using the new lustre-2.4.3 client.

nmngas01 was apparently injesting a 30GB VLBA file when I rebooted it.
Because of this it tried to move it to the bad-files area which is on
/ and does not have enough space for a 30GB file.  Stephan moved this
file to Volume30 for now and John will look into putting it where it
really belongs.  In the future, the procedure should be to stop
injestion before rebooting these machines.

mctest and mcahmmer came up without issue but last time we upgraded
them I think Bryan had to start some process.  Does anyone know what
that might be?


On Feb 21 08:30, K. Scott Rowe wrote:
}Two weeks ago, I switched all the machines that mount the EVLA Lustre
}system to the new EVLA lustre system.  I did not upgrade the version
}of Lustre on these machines.  I specificly wanted that to be a
}seperate step in case we had to roll back and because this step takes
}some time.  Now I am ready to upgrade the version of Lustre on these
}machines.
}
}This process involves:
} - Upgrade mchammer and mctest from RHEL6.7 to RHEL6.8. ~2 hours
} - Upgrade the Lustre client on nmngas01 and nmngas02 ~30 minutes
} - Upgrade the Lustre client on cbe-master and cbe-node-* ~30 minutes
}
}The whole process should take about 3 hours.  It will require a reboot
}of every machine listed above.
}
}I would like to do this on Feb. 22 starting about 9am and probably
}finishing around noon.  Please let me know if this is unaccaptable for
}any reason.
}
}Thanks
}
}_______________________________________________
}ngas-alerts mailing list
}ngas-alerts at listmgr.nrao.edu
}https://listmgr.nrao.edu/mailman/listinfo/ngas-alerts