Hey,

We are using Zero Downtime backup in the DIY backup in our Bitbucket DC in AWS and we have run into an issue with the locking of the EBS just before the EBS snapshot is taken.

Between the freeze and unfreeze commands there is the actual snapshot command issued which is more or less instant. But there is also additional aws cli commands issued to grab info about the volume to snapshot etc. These cli commands takes time that increases the time the file system is locked.

In our large instance we ran into the issue that one CI/CD pipeline started just when the filesystem was locked and requested a very high amount of git ops. These ops could not be processed due to the filesystem being frozen. 

I have clocked the freeze time in our system to be 5 s on an average and I think the aws cli commands are responsible for most of this. Would it be possible to move these commands outside of the freeze period, perhaps?

Cheers,

// Svante