I recently started with a company that relies on an aging NFS server that handles files for thousands of sites. This server, integral to our NGINX setup, has seen years of operation and is starting to run out of space, now nearing its 10% free limit. We've tried deleting old snapshots, but we need to keep a month's worth for compliance. The system also uses ZFS without mirroring, leading to concerns about potential single points of failure.
Given these constraints, I'm considering provisioning a new, larger NFS system with three instances across different availability zones, leveraging GlusterFS for replication. My goal is to maintain redundancy and reliability, but I'm unsure if GlusterFS is the best choice for our situation, especially with heavy read/write usage involving lots of small files. I'm also curious about block replication options and the ability to keep files synced across multiple nodes. Does this strategy sound reasonable, or am I overlooking something essential?
3 Answers
It really boils down to cost-effectiveness and operational efficiency. Have you compared the expenses of maintaining a GlusterFS cluster against using AWS Elastic File System (EFS)? EFS offers scalable storage with room for tons of data, and it integrates nicely with AWS services like backups. Just make sure you're ready to handle the performance tuning if you go with GlusterFS, as you'll need to keep an eye on how it's set up, especially with replication across regions.
This whole situation definitely falls under the DevOps umbrella! It’s fascinating to hear about the complexities of managing legacy systems while trying to modernize. Your idea of using three instances is smart, especially to mitigate risks. Just keep in mind that while you’re on the right track, the complexity of syncing files in an old setup can increase with the number of nodes. Be prepared for some growing pains, but it sounds like a plan that's worth pursuing!
If you're avoiding EFS to steer clear of vendor lock-in, GlusterFS is definitely a solid alternative for an NFS-style setup. It can easily mesh sync between three nodes, letting you failover by changing DNS settings if needed. Just a heads-up, remember that any extra disks you add won't stick around after a restart unless you build them into your launch template. Also, don't forget to test failover scenarios ahead of time to avoid surprises when things get hectic.
Related Questions
Can't Load PhpMyadmin On After Server Update
Redirect www to non-www in Apache Conf
How To Check If Your SSL Cert Is SHA 1
Windows TrackPad Gestures