How to obtain a package:
1. Minion locates upstream site with package
2. Minion wget's it
3. Minion replicates it out to your various package repositories
4. Minion says it's done
5. Your installation fails
6. You discover minion failed to get the URL correct, and instead has replicated out a copy of a 404 page called the package.
7. Smack minion. Bad minion. No biscuit.
Tuesday, November 11, 2008
Thursday, October 23, 2008
Choices
Your VM host is out of space. You can't add more disk to it, as the chassis is full. There's no SAN.
You have two choices:
(a) use a spare VM host of a similar spec and migrate some machines to it
(b) break the RAID-1 on the exsting host, and run it with no disk redundancy
If you chose (a), you're wrong. The obvious answer is (b).
Right?
You have two choices:
(a) use a spare VM host of a similar spec and migrate some machines to it
(b) break the RAID-1 on the exsting host, and run it with no disk redundancy
If you chose (a), you're wrong. The obvious answer is (b).
Right?
Tuesday, August 12, 2008
Networking is hard
You have a new VM to build. It has three IP addresses on the same network. Naturally, you set up the name of the machine in the DNS with one of them, and have suitable application names for the others. Maybe you CNAME things as well. And you'll need a virtual NIC too.
That's why you do this:
# host myvmname
myvmname.domain has address 10.10.1.20
myvmname.domain has address 10.10.1.22
myvmname.domain has address 10.10.1.21
And why you don't provide any other names.
And why you give the VM three virtual NICs, one for each IP address. All of them on the same network.
Networking is hard.
That's why you do this:
# host myvmname
myvmname.domain has address 10.10.1.20
myvmname.domain has address 10.10.1.22
myvmname.domain has address 10.10.1.21
And why you don't provide any other names.
And why you give the VM three virtual NICs, one for each IP address. All of them on the same network.
Networking is hard.
Wednesday, July 16, 2008
Requirements. You don't has them.
User: Can you copy thisapplog to othermachine please?
Me: Where do I find thisapplog at the moment?
User: I don't understand
Me: What machine is thisapplog on?
User: I don't know
Me: ...
Resisting the urge to put a "copy" of the log file containing nothing but:
[Wed 16 Jul 2008 04:43:56] error: you fail at specifying logs
Me: Where do I find thisapplog at the moment?
User: I don't understand
Me: What machine is thisapplog on?
User: I don't know
Me: ...
Resisting the urge to put a "copy" of the log file containing nothing but:
[Wed 16 Jul 2008 04:43:56] error: you fail at specifying logs
Friday, June 27, 2008
Friday 5pm
It's 5pm on Friday. You are a few hours out from a massive migration of data between SANs. The planning for this has been rough, but it is starting to look sane.
That is until you discover right then that no-one has mentioned before the size of the disks you'll be getting from the new SAN is different to the old one.
It's not like that would be an important detail.
This is Friday 5pm for sysadmins.
That is until you discover right then that no-one has mentioned before the size of the disks you'll be getting from the new SAN is different to the old one.
It's not like that would be an important detail.
This is Friday 5pm for sysadmins.
Tuesday, June 17, 2008
Gah!
No, I cannot burn your 50GB of data to 'a DVD'. Even if my regulation issue company brick had a blueray disk burner, I'd still have to get the data back to it through may layers of VPN and NAT that at some points only manages to pass data at slow internet speeds. This is just a silly idea, honestly. Buy a real backup solution.
Wednesday, May 7, 2008
The Trouble Ticket
When I first started a job at a recent employer, I was quickly assigned The Ticket. This ticket had been in the system since time immemorial, and had been passed around all of the administrators in the team. The comment history scrolled for pages and pages.
On the surface, it didn't really look all that bad. There was a small development company who were dissatisfied with their current revision control system and practices, and who wanted a new system put in place and some training in how to use it. They were very happy with open source solutions and they were using a revision control system old enough there were a lot of scripts to migrate from it to almost any of the newer systems.
Apart from everyone having a different opinion on which system to migrate them to, I was fairly hopeful it was going to be an interesting project, or at the very least not terribly difficult.
To start with, I brought the discussion around to the customer's current usage patterns.
'Well, one of the problems that we have is disk space on the local developer workstations. We don't want a checkout of the code on every system, so we check out the code onto a file server and then everyone edits it from there'
This sounded extremely odd to me, but I assumed naively that people had their own checkouts in their own shell accounts.
'Oh no, we share the code out on a Samba share, and everyone maps it as a drive on their workstations'
Oh no.
It turns out that the entire office were mounting the same samba share read/write all using the same username and password, and all editing the files as the same user in the revision control system. There was absolutely no ability to figure out who was editing what, and conflicts happened regularly. When they did, work would grind to a halt for hours while everyone tried to figure out a compromise. What was worse, the disk space problem that had prompted this odd solution was mostly a factor of their lack of understanding of how their revision control system worked. Not everyone needed a full checkout of the repository but they hadn't seemed to yet figure out that you didn't need to pull the whole thing.
I made my escape quickly and when I got back to the office I ignored The Ticket.
Until the next new Sysadmin started, and then I assigned it to him.
On the surface, it didn't really look all that bad. There was a small development company who were dissatisfied with their current revision control system and practices, and who wanted a new system put in place and some training in how to use it. They were very happy with open source solutions and they were using a revision control system old enough there were a lot of scripts to migrate from it to almost any of the newer systems.
Apart from everyone having a different opinion on which system to migrate them to, I was fairly hopeful it was going to be an interesting project, or at the very least not terribly difficult.
To start with, I brought the discussion around to the customer's current usage patterns.
'Well, one of the problems that we have is disk space on the local developer workstations. We don't want a checkout of the code on every system, so we check out the code onto a file server and then everyone edits it from there'
This sounded extremely odd to me, but I assumed naively that people had their own checkouts in their own shell accounts.
'Oh no, we share the code out on a Samba share, and everyone maps it as a drive on their workstations'
Oh no.
It turns out that the entire office were mounting the same samba share read/write all using the same username and password, and all editing the files as the same user in the revision control system. There was absolutely no ability to figure out who was editing what, and conflicts happened regularly. When they did, work would grind to a halt for hours while everyone tried to figure out a compromise. What was worse, the disk space problem that had prompted this odd solution was mostly a factor of their lack of understanding of how their revision control system worked. Not everyone needed a full checkout of the repository but they hadn't seemed to yet figure out that you didn't need to pull the whole thing.
I made my escape quickly and when I got back to the office I ignored The Ticket.
Until the next new Sysadmin started, and then I assigned it to him.
Subscribe to:
Posts (Atom)