https://samirtech.org/tags/lessons-learned/ Recent content in lessons-learned on Samirtech Hugo -- gohugo.io en-us Wed, 16 Sep 2026 03:00:00 +0100 https://samirtech.org/posts/what-i-would-do-differently-if-i-started-again/ Wed, 16 Sep 2026 03:00:00 +0100 https://samirtech.org/posts/what-i-would-do-differently-if-i-started-again/ The most important lesson is that I should have slowed down before selecting the TrueNAS reset option while moving home. I did not read carefully enough, and the result was a wipe of the system hosting my photo library. If I started again, I would change the following. I would separate data from experimentation I would make the consequences of destructive operations obvious before pressing the button. I would export configuration, confirm snapshots and check the target dataset before resetting or rebuilding anything. <p>The most important lesson is that I should have slowed down before selecting the TrueNAS reset option while moving home. I did not read carefully enough, and the result was a wipe of the system hosting my photo library.</p> <p>If I started again, I would change the following.</p> <h2 id="i-would-separate-data-from-experimentation">I would separate data from experimentation</h2> <p>I would make the consequences of destructive operations obvious before pressing the button. I would export configuration, confirm snapshots and check the target dataset before resetting or rebuilding anything.</p> <h2 id="i-would-document-recovery-while-everything-worked">I would document recovery while everything worked</h2> <p>I would record the TrueNAS dataset layout, Immich paths, database location, backup roots, B2 details and restore commands before an emergency. Recovery knowledge is hardest to reconstruct after a failure.</p> <h2 id="i-would-protect-keys-first">I would protect keys first</h2> <p>I would keep TrueNAS encryption keys and Duplicati/B2 passphrases in a password manager, encrypted offline archives and a second safe location. The keys would be labelled clearly so I knew which system and dataset each one belonged to.</p> <h2 id="i-would-test-restores-regularly">I would test restores regularly</h2> <p>I would restore a sample of old photos and videos, rebuild a test database and check hashes on a schedule. I would not wait for a disaster to discover that an old backup was incomplete or depended on missing software.</p> <h2 id="i-would-keep-multiple-recovery-paths">I would keep multiple recovery paths</h2> <p>The recovery worked because several imperfect sources overlapped: Voyager, Backblaze B2, Duplicati information, database dumps and older files. I would design those paths deliberately instead of discovering them after the loss.</p> <h2 id="i-would-not-delete-old-sources-too-quickly">I would not delete old sources too quickly</h2> <p>After a migration, I would keep the source untouched until I had checked counts, sizes, old years, originals, videos, albums and database health. A migration summary is not the same as a verified restore.</p> <h2 id="i-would-make-the-plan-boring">I would make the plan boring</h2> <p>The best backup system is not the cleverest one. It is the one that runs automatically, keeps more than one copy, stores its keys separately and has a restore procedure I can follow when tired and worried.</p> <p>I was fortunate to recover my photos, including family memories going back to 2009. I do not want the next recovery to depend on fortune.</p>