I’m refreshing a number of CircleCI build projects, and finding that builds that used to take 3-4 minutes are now taking 13m+. It looks like they are particularly getting stuck in network operations, e.g. apt-get upgrade in Ubuntu.
Anyone else getting this? If so, is it new in the last hour or two, or has it being going on all day, etc? I’ll update here if and when it goes away.
Ah, our ops provider says that archive.ubuntu.com is struggling for him in AWS, which may well be the same problem I have here.
Tip received from our ops provider:
sed -i 's#archive.ubuntu.com/ubuntu#ftp.heanet.ie/pub/ubuntu#g' /etc/apt/sources.list
sed -i 's#security.ubuntu.com/ubuntu#ftp.heanet.ie/pub/ubuntu#g' /etc/apt/sources.list
And then swap the target to the closest/fastest mirror at https://launchpad.net/ubuntu/+archivemirrors.
I’m trying this now with a US mirror (Enzu) as being most suitable for CircleCI, it is working fine even when building in Docker locally (in the UK).
Hi @halfer,
Not sure how do you apply this fix in the context of a machine executor.
Do you kill -9 the currently running apt-get, edit the sources file then re-execute apt-get update?
As I mentioned on the other thread, you are bumping into two issues at once. The machine executor seems to have a problem where it takes a long time to unlock the local repo files. I seem to recall there is a fix on another thread here - it basically involves waiting in a loop until the local repo is unlocked by the OS.
Edit to add link:
@halfer and anyone else reading this, I added another solution to that thread about the Could not get lock /var/lib/apt/lists/lock error.
Thank you @katiekeel - that problem has come up a few times on the forum, and isn’t helped by Ubuntu’s networking woes 
.
Yes, I’m managing to have both that problem and the networking problem simultaneously today. 
This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.