Hacker News new | ask | show | jobs
by alephnerd 757 days ago
> In VM's case it is not that long really

It is. You have teams critical apps running and you do not want to cause downtime for customers (internal or external).

I've seen this saga happen dozens of times and it is a very delicate and intricate process that requires a lot of planning.

A failed migration can become news, for example - https://www.bloomberg.com/news/articles/2023-11-15/ubs-says-...
1 comments
prmoustache 757 days ago
It only is because companies usually don't staff to be ready for large migrations. They should.
link
icedchai 757 days ago
They'll be sitting around twiddling their thumbs most of the time. Seems like something to be contracted out when necessary.
link
alephnerd 757 days ago
Why should I hire an additional 50 Infra Engs who are underutilized when I can use that to hire 50 additional SWEs or Salespeople to expand my TAM?
link
prmoustache 757 days ago
Because migrations should happen all the time. If things are done right requires the same effort to patch and have downtime of all your systems than it is to migrate from one hypervisor to another. I've migrated tons of systems, sometimes p2v and v2v makes it simpler/faster but more often than not it is more useful and as fast to reinstall system then redeploy app (you can often use that time to migrate to major OS release) and it is dead easy to do a rollback if you encounter issues in migration, stop the new one, restart the old one.

A company that is not staffed to migrate all its servers in less than 3 months is not staffed to have the most minimum level of IT security.

link
icedchai 757 days ago
The reality is that in many organizations, nobody wants to update or reboot anything unless absolutely necessary. I once took a contract job at a company that wanted me to migrate and upgrade some Linux VMs running on a 5+ year old distro. This was a decent sized, established company, not a startup (1000's of employees, 100 million+ revenue)

The existing VMs had not been updated or rebooted in years. Several had 1200+ day uptimes! These systems were accessible from the Internet and regularly used by customers for an important, highly specialized product. The only reason they were even bothering was they had failed some security scan and nobody there could even understand how to use SSH.

The minimum level of IT security is apparently barely anything.

link
alephnerd 757 days ago
> Because migrations should happen all the time

Hypervisor migrations? Complete DC migrations? If you are doing that every single quarter you are doing something VERY wrong (or you should migrate to the cloud).

link
prmoustache 757 days ago
You don't have to do these migrations, but you should be able to do them because basic security maintenance of systems require regular patching, reboots, downtimes, etc. Setting up a new VM and deploy the apps/services takes minutes, not days.

The infra has to be setup to support downtime of systems (not all at once obviously).

link