I was sick of staring at filesystem translator configs.
That did not mean the work was going badly. The first Gluster proof of concept was complete and already serving a customer workload. The next version would combine distributed and replicated storage across four nodes.
Reality Check sold managed infrastructure. We needed storage that could handle our own systems and customer systems without turning every growth step into another expensive box or another emergency migration.
I had already tested older Gluster options and found them short on the features we needed. The newer version looked more promising. I was considering Solaris and ZFS underneath it and wanted several weeks of performance data from different layouts before making a larger call.
The immediate goal was not a grand storage-platform launch. It was a working four-node cluster, measured under real conditions, with the caveats written down.
Other work was piling up around it. A customer needed a network proposal. The firewalls needed attention. We still did not have the VPN setup I wanted. Layer-three switch upgrades were waiting, and virtual machines needed to move onto the company's defined hypervisors.
The previous storage headaches would also shrink once the VMware migration finished. That mattered because it changed the priority. A problem affecting the business every day deserved a different response from an annoyance that could be measured for another few weeks.
I gave myself two more days for the replicated cluster. After that I planned to stop touching Gluster for at least a day and clear some of the network work.
This was the less glamorous version of a new CTO role. The title said executive. The work bucket said switches, VPN, filesystems, hypervisors, customer proposals, and a backlog that needed another three months of stories.
The proof of concept worked. I still needed to look at something else.