Operations analyst (daily operations runbook)

First job, obtained through referral. The operations unit had three people contracted as operations analysts; the hire was made as an operations technician, on the same activity and a lower pay grade. Arrival came with web development experience and no knowledge of operations: Linux on a daily basis, but no shell scripting, no scheduling, no batch chain.

The day was entirely scripted. An operations runbook set out some thirty tasks to be performed every day, at the same times and in the same order, first in pairs and then alone. Before nine: reading the overnight batch status, identifying failures, reporting by email, analysing and restarting where possible, raising a vendor ticket otherwise; starting the transactional system and running a liveness test; checking backup logs, exporting to tape through a Tivoli menu, then physically moving the tapes off-site to a branch office a kilometre away.

Nine o'clock, a standing call with the development teams, operations and the business owners on the night just past. Ten, manual launch of the direct debit and transfer flow. Four in the afternoon, a second call, followed by the evening request — a file sent by the development teams listing the specific jobs to be inserted into the night's schedule, in production as well as outside it. Five, collection and consolidation of the flows under the NOEMIE and IRIS B2 standards. Six, shutdown of the acceptance transactional system and manual launch of its batch chain. Half past seven, shutdown of the production transactional system, flash copy, backups, then the batch chain.

To this were added the monitoring and restarting of CFT flows through to the application, and the installation of vendor releases and fixes, run after the transactional system closed, on scripts supplied by the database administrators. Production machines ran on AIX. A great deal of manual action, and a great deal of command-line work.

The disaster recovery plan was exercised one week a year: the systems team travelled to the hosting provider in Paris with the off-site tapes, brought the servers back up, and operations launched the jobs remotely to verify that the whole ran. The team grew from three to six in 2015, when the operations staff of the absorbed entity joined.

A Unix course taken early in the post opened up scripting and set off a series of automation projects that would fill the five years: each one freed time, and that time funded the next. The role had outgrown its contractual title — system administration, production industrialisation, project management — without the reclassification that had been requested. A move to Montpellier and the wish for a different setting brought the period to an end in 2017.