Comments on: My server uptime

By: Jack Hughes

Jack Hughes — Fri, 28 Aug 2009 15:39:09 +0000

As an update, since the retirement of our Java based e-commerce system the load has come down significantly: load average: 0.15, 0.16, 0.11 as opposed to 1.15, 1.23, 1.49

By: Tom T

Tom T — Mon, 13 Jul 2009 23:04:10 +0000

Solaris 8

Yes it does seem that patching has been a tad neglected on that box but it would seem a shame to destroy that uptime now

By: Jack Hughes

Jack Hughes — Fri, 10 Jul 2009 13:20:53 +0000

@Tom – wow, what OS are you running there?

By: Tom T

Tom T — Fri, 10 Jul 2009 12:37:42 +0000

Nearlly 4 years…

***:/> uptime
1:32pm up 1443 day(s), 2:54, 3 users, load average: 0.79, 1.43, 1.32

By: Jack Hughes

Jack Hughes — Mon, 29 Jun 2009 08:13:14 +0000

@Jamie – the bottleneck is a rather large Java based e-commerce system that will be retired in the next couple of weeks. It hammers the server 24/7 whatever the traffic load… It is amazing the server has coped with it so well.

By: Jamie Isaacs

Jamie Isaacs — Fri, 26 Jun 2009 17:48:39 +0000

I guess kernel patches containing security updates are less important than the all knowing uptime metric. Why not place the uptime statistic on individual “services” instead of an archaic server uptime? Then establish a high availability layer with something like heartbeat. Then you are free to reboot servers for upgrades as long as each “service” have a failover path. The “service” in this case could be anything from apache to dhcp.

That is a relatively high 15 min load average by the way. Have you found the bottleneck?

By: Dave

Dave — Fri, 26 Jun 2009 17:24:55 +0000

Why would someone post to the world about doing a terrible job keeping up with security patches?

By: Nico

Nico — Fri, 26 Jun 2009 17:23:13 +0000

18:44 nico@jaguar ~> uptime
6:44PM up 755 days, 3:20, 1 user, load averages: 0.00, 0.00, 0.00

(Running FreeBSD 4.11)

By: Nico

Nico — Fri, 26 Jun 2009 17:22:41 +0000

18:44 nico@jaguar ~> uptime
6:44PM up 755 days, 3:20, 1 user, load averages: 0.00, 0.00, 0.00

By: David Mackintosh

David Mackintosh — Fri, 26 Jun 2009 16:55:59 +0000

Ah, the uptime DSW. There used to be an internet project somewhere that let you install a client that would send your uptimes on a regular basis to a central host which would then rank everyone by uptime.

I used to think uptime was king… until I noticed that even modern unix OSs (or more accurately, the applications that end up running on them) can leak or zombify or whatever… over time this junk just accumulates and you need a reboot to flush it out.

However, there’s a better reason:

“Unix administrators are slightly more free, key word being slightly. Depending upon the vendor, they usually patch quarterly. Gone are the days of the 365+ day server uptime. In fact, if you run into a Unix admin boasting of 365 day server uptime, feel free to thank him for letting you know that his systems are missing several quarterly security patch releases, therefore letting the free world know that his systems are vulnerable to whatever ailment(s) were patched in the past year.”

…from http://www.sysadmin-network.com/profiles/blogs/reboot-with-pride

…that said, I have one system with 500+ days of uptime.

[root@saturn ~]# uptime
12:54:35 up 560 days, 42 min, 1 user, load average: 0.00, 0.01, 0.00

As it happens, it’s getting patched and rebooted on Monday.