I'm starting a new role where I have to deal with an aging NFS server that serves thousands of websites. This server, which relies on ZFS for its storage, is approaching critical low space—down to 10% free space—and deleting old snapshots hasn't helped as we need to retain one month of archives. I can't install updated tools or create a new image due to compatibility issues with the existing system. I'm considering setting up a new NFS server that's larger and consists of three instances to act as replicas using GlusterFS. My idea is to place the instances in different availability zones, with two on standby in case one fails. I'd like to know if this approach of using GlusterFS for replication is solid, especially given the structure of many small files within numerous deep folders. Are there other better tools for this replication? Also, can GlusterFS manage bidirectional replication among the instances? I want to avoid vendor lock-in with AWS solutions, so I'm looking for the best route to take.
3 Answers
Before diving into a clustered NFS with GlusterFS or Ceph, it might be worth comparing the costs of those options against using AWS EFS. EFS offers tiered storage and an astonishing 8 Exabytes of capacity, plus it's integrated with AWS Backup and supports encryption. For scaling, keep in mind you’ll need to calculate the operational costs of managing multiple instances versus using EFS—don’t forget to factor in patching and maintenance as well. If you do opt for GlusterFS, be prepared to do some performance tuning on your storage volumes and configure replication carefully. There are plenty of options out there, so maybe do a proof of concept before making a final decision!
Isn't this a part of DevOps? If you're concerned about vendor lock-in with EFS, going with GlusterFS for traditional NFS-style replication is a great choice. It allows for full mesh sync between three nodes and can easily handle failover through DNS changes. Just keep in mind that any disks you add manually won't survive a disaster if your infrastructure goes down since the launch template only tracks main EBS volumes. Test your failover plan thoroughly before an actual situation arises; dealing with it late on a Friday is never a good time!
Absolutely, this falls under DevOps! The challenges you're facing are pretty common, and this setup could definitely work if planned properly. You can achieve synchronous replication with GlusterFS and keep all instances updated. The key is to make sure that you check the performance and tweak the configuration as needed. If you haven't already, give a solid thought to testing your failover connection now and not wait until the last minute—it will save you a lot of trouble later!
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