Concern about "apt-get update" in production server

Question

I have a server which is getting a bit outdated as I haven't updated it since a few months. I wanted to try cron-apt but first wanted to do it manually, so I issued a

sudo apt-get update

as a first step. Everything went ok but when I checked my apache error logs I got some messages like these:

[error] [client xxx.xxx.xxx.xxx] exception 'MongoCursorTimeoutException' with message 'cursor timed out (timeout: 30000, time left: 0:0, status: 0)'

from users who were connecting to my server at that moment.

Please note that I haven't issued a apt-get upgrade yet, but I got a bit concerned that just updating the package db would affect some users, specially if I cron this to say once a day.

Only workaround I can think of is doing this at a off-peak time, but besides that obvious assumptions, is there any other tip you could give?

Describe the OS distribution, version, information about the hardware environment, etc. — Michael Hampton, May 22 '13 at 21:36

score 4 · Accepted Answer · answered May 22 '13 at 21:39

4

This is why you should have a backup server. Or at least a test environment that can pretend to be a server in a pinch.

Upgrade the backup server.
TEST the backup server.
Fail over to the backup server.
Upgrade the main server.
TEST the main server.
Fail back to the main server.

As for apt-get update messing with other stuff, it shouldn't be an issue but if you are on a low-power server (eg micro VM instance) it can run up CPU load and memory use which could slow down other programs.

answered May 22 '13 at 21:39

DerfK

19,493
2
38
54

My bet is that is all about the CPU/Memory Load. Rebuilding the package databases is a relatively intensive process. – Zoredache May 22 '13 at 22:19
1

This is also why you shouldn't run anything production on a t1.micro. – ceejayoz May 23 '13 at 00:41

Concern about "apt-get update" in production server

1 Answers1