Manual testing has hit several cases where a node that had run an online restore was running but if stopped was unable to restart. One one case manifest entries used only during startup were broken, and in #120253 we observed mismatched file sizes. We should add a roachtest that stops and restarts clusters during and after the download phase to ensure that the we test such cases.
Manual testing has hit several cases where a node that had run an online restore was running but if stopped was unable to restart. One one case manifest entries used only during startup were broken, and in #120253 we observed mismatched file sizes. We should add a roachtest that stops and restarts clusters during and after the download phase to ensure that the we test such cases.
Jira issue: CRDB-36566