Merge branch 'db-okr-scoring' into 'master'

Added DB OKR retrospective for Q4 of 2017

Closes #1880

See merge request gitlab-com/www-gitlab-com!9264

source/okrs/2017-q4/index.html.md

@@ -167,10 +167,10 @@ title: "2017 Q4 OKRs"
     * Generate a project plan for the GCP migration and get approved by EJ and Sid
     * Execute milestone 1 of the GCP migration plan by Dec 15
   * Database Lead
-    * Demo restore time < 1 hour
-    * Solve 30% of the schema issues identified by Crunchy
-    * Database Uptime 99.99% measured in Prometheus
-    * SQL timing under 100ms for Issue, MR, project dashboard, and CI pages measured in Prometheus
+    * Demo restore time < 1 hour => postponed until the GCP migration has been completed
+    * Solve 30% of the schema issues identified by Crunchy => 6.5% done (2 out of 30)
+    * Database Uptime 99.99% measured in Prometheus => Done!
+    * SQL timing under 100ms for Issue, MR, project dashboard, and CI pages measured in Prometheus => Improving the 99th percentile has proven to be very difficult, but progress being made.
   * Director of Security
     * Strong security for SaaS and on-premise product. Top 10 actions from risk assessment done and actions for top 10 risks started.
     * HackerOne bug bounty program. Implemented and bounties awarded.
@@ -240,7 +240,7 @@ title: "2017 Q4 OKRs"
   * Hire 2 senior developers
 * Discussion: Hire 2 developers
 * Gitaly: Hire a developer
-* Database: Hire a database specialist
+* Database: Hire a database specialist => Done! Starting end of January 2018
 * Director of Quality
   * Hire a test automation lead
   * Hire 3 test automation engineers
@@ -293,7 +293,24 @@ title: "2017 Q4 OKRs"
     * Continue collaborating with FE on UI Repository and UX Backlog cleanup
     * Push harder for significant iterative UX improvements in each release
     * Anticipate holiday season's influence on ability to deliver
-* Database Lead
+* Staff Developer, Database
   * GOOD
+    * Despite the large OKR we managed to solve a lot of performance issues.
+    * We improved the team workflow by using issue boards more actively and by having a weekly database meeting.
+    * We managed to add health / uptime monitoring to Prometheus / Grafana, allowing us to see how the database health changes over time. This is based on the number of alerts sent out, not the uptime of the database.
+    * We managed to hire a 3rd database specialist.
+    * We rewrote the GitHub importer from scratch, resulting in _much_ better performance.
+    * We wrote a (popular) blog post about scaling the database: <https://about.gitlab.com/2017/10/02/scaling-the-gitlab-database/>
+    * We managed to optimise retrieving CI pipeline statuses, which used to execute _very_ slow SQL queries.
   * BAD
+    * We added far too much work to the Q4 OKR, resulting in us only being able to complete a small portion of the planned work.
+    * We didn't take the summit into account when planning the OKR.
+    * One database specialist was unavailable for a few weeks due to having to move to a different apartment. This lead to a reduction in productivity of the team as a whole.
+    * There were too many issues that required the help of others, some of these were not worked on for several weeks.
+    * We estimated we'd be able to complete 30 schema issues, but only ended up completing two of them.
+    * 10.3 had a few bad migrations causing trouble.
   * TRY
+    * Schedule more well defined issues for an OKR so we can actually solve them.
+    * Make it harder to introduce performance problems (planned for Q1 of 2018).
+    * Assign database specialists to specific areas instead of having them take care of everything (<https://gitlab.com/gitlab-com/infrastructure/issues/3139>).
+    * Delegate more work to the other teams so database specialists don't have to do so much one their own.