|
Written by Todd Green
|
|
Thursday, 17 August 2006 08:36 |
While an outside team was working on the fire suppression system in the main machine room, MEB 3555, they somehow caused a cataclysmic power failure in the room.
The SoC servers and network equipment is on UPS power, but it appears that Netcom's equipment in the room is not - so we temporarily lost connection to the outside world. (This has been rectified and Netcom's equipment is running on our UPS until their UPS is installed.)
Power has been restored and things should be back to normal. The fire suppression team is backing off from the project until they can determine what caused this failure during testing. |
|
Last Updated on Monday, 18 September 2006 03:34 |
|
|
Written by Todd Green
|
|
Thursday, 10 August 2006 02:50 |
Summary:
The OS upgrade on the filer caused a bug where one CPU hit 100% utilization and slowed CIFS (Windows) file operations. It also broke our quota tools. This has been worked around and things are back to normal. Actually better than normal as csquota will now cache your index so subsequent lookups will happen almost instantaneously.
Details:
There were performance issues unless default quotas were set for all file systems and groups. Since we didn't use group quotas, turning them on broke our quota tools. (Note that groups are set to have infinite quota, so it shouldn't impact your disk usage). They also bloated the SNMP tables since we have quite a few groups. There were some other changes to how they stored the data as well. csquota has been modified to handle groups, the new storage format, and cache your lookup index so that after being run the first time, it can go directly to your entry in the table instead of having to walk 18,000 or so entries. The cache will be valid unless the filer needs to be rebooted or the quota system restarted which is infrequent, but you may notice that csquota runs slowly (30 seconds or so) the first time until it can revalidate the cache. We've also speed up reporting research quotas which used to take around 20 seconds and now runs in a second or two. Finally, a reminder that csquota has supported a -h flag for awhile now to get "human readable" formatting of the output. |
|
Last Updated on Monday, 18 September 2006 03:34 |
|
|
|
|
|
Page 16 of 16 |