PVC backed by EBS on EKS (running a script on CF to mkfs with high inode count)?
Cluster was running loki with just 3 days worth of log retention and ran out of inodes. Had changed config to use bigger blocks so should be fine for new ones but need to take care of already stuck loki.
By default EBS was mounted with 650,553 inodes over 10gb.
So something like this but k8s:
https://stackoverflow.com/questions/35574281/increase-the-number-of-inodes-on-aws-ebs
Push those to cloudwatch and start removing them from the oldest
Обсуждают сегодня