1. 10 Aug, 2020 4 commits
  2. 07 Aug, 2020 5 commits
    • Sami Hiltunen's avatar
      report only read-only repositories by default in dataloss · 9d0937cc
      Sami Hiltunen authored
      `dataloss` command currenly display every outdated repository on a
      virtual storage by default. While outdated secondaries should also
      be fixed, they do not require immediate administrator action as they
      do not affect the writability of the repository. Secondaries might also
      be outdated only temporarily until a replication job brings them up to
      date.
      
      Outdated primary requires administrator action to enable the repository
      for writes again. This is currently a permanent state until administrator
      takes action. To reduce the noise from the tool, this commit only displays
      read-only repositories by default. Listing outdated writable repositories is
      still supported via the `-partially-replicated` flag.
      9d0937cc
    • Patrick Steinhardt's avatar
      Merge branch 'smh-configurable-dequeue-batch' into 'master' · 30098071
      Patrick Steinhardt authored
      Configurable replication queue batch size
      
      See merge request !2450
      30098071
    • Sami Hiltunen's avatar
      configurable replication queue batch size · 4bb05403
      Sami Hiltunen authored
      Adds a configuration option for controlling the replicator's job
      dequeueing batch size.
      4bb05403
    • Pavlo Strokov's avatar
      Merge branch 'pks-disable-reference-transaction-hook' into 'master' · 6823bffa
      Pavlo Strokov authored
      Remove reference-transaction hook
      
      Closes gitlab#233964
      
      See merge request !2454
      6823bffa
    • Patrick Steinhardt's avatar
      git-hooks: Remove reference-transaction hook · 19e2caa3
      Patrick Steinhardt authored
      There's a bug in the reference-transaction implementation in git-core
      which will cause interleaving calls to the reference-transaction hook to
      execute another hook by accident. While a fix has been proposed
      upstream, the earliest it may be available is with Git v2.28.1. So let's
      remove our reference-transaction hook for now to not expose this broken
      behaviour.
      19e2caa3
  3. 06 Aug, 2020 9 commits
    • Ethan Reesor's avatar
      GetArchive: Support path elision · 9629a924
      Ethan Reesor authored
      9629a924
    • Patrick Steinhardt's avatar
      Merge branch 'pks-praefect-info-ruby-sidecar' into 'master' · d6ded95a
      Patrick Steinhardt authored
      Fix connecting to Praefect service from Ruby sidecar
      
      Closes #3017
      
      See merge request !2451
      d6ded95a
    • Patrick Steinhardt's avatar
      rubyserver: Resolve Praefect server address before re-injecting it · c4e92421
      Patrick Steinhardt authored
      When an RPC call's metadata contains Praefect server metadata, then this
      metadata may not be used to connect to Praefect immediately. Most
      importantly, it may contain an address that we cannot use in the current
      context to reach back out to Praefect, e.g. as it may contain the
      wildcard address "0.0.0.0". As such, we usually resolve Praefect server
      information before actually using it by using the current context's peer
      information, which contains information about how the peer connected to
      us. Thus, by replacing the passed Praefect server address with the
      actual one we should now be able to reach out to it.
      
      While we do the above for "normal" code, we don't currently do so when
      proxying RPC calls to the Ruby sidecar: we'll just pass on metadata
      as-is. As the Ruby sidecar now has Gitaly as its peer and not Praefect,
      it's impossible to resolve the address at that point in time.
      
      Fix the issue by resolving in the proxying code directly such that the
      Ruby sidecar will receive correct information and won't have to resolve
      itself. Add a test to verify we can connect to catch-all TCP addresses
      and others.
      c4e92421
    • Patrick Steinhardt's avatar
      Merge branch 'smh-revert-dequeue-batch-size' into 'master' · 2a5be332
      Patrick Steinhardt authored
      Revert "dequeue one event per worker at a time"
      
      See merge request !2452
      2a5be332
    • Sami Hiltunen's avatar
      Revert "add missing changelog entry" · ceb419ce
      Sami Hiltunen authored
      This reverts commit 07d6568f.
      ceb419ce
    • Sami Hiltunen's avatar
      Revert "dequeue one event per worker at a time" · 52102245
      Sami Hiltunen authored
      This reverts commit 20d61879.
      52102245
    • Patrick Steinhardt's avatar
      praefect: Split up creating and injecting Praefect server · b827ca57
      Patrick Steinhardt authored
      We currently only expose a single function to inject Praefect server
      information into a context, which takes as argument the Praefect
      server's configuration. This is inflexible and doesn't allow us to
      re-inject information from connection information obtained via other
      means.
      
      Fix this by splitting up concerns of those functions.
      `PrafectFromConfig()` creates a new structure from the Praefect
      configuration, while `Inject()` is a receiver which injects the
      structure into a given context.
      b827ca57
    • Pavlo Strokov's avatar
      Merge branch 'ps-enable-features' into 'master' · 9e5dfd98
      Pavlo Strokov authored
      Feature flags enabling for tests.
      
      See merge request !2425
      9e5dfd98
    • Pavlo Strokov's avatar
      Feature flags enabling for tests. · a1731804
      Pavlo Strokov authored
      It is decided to go with all features enabled by default behaviour
      for tests. This should help us identify potential problems with
      feature flag combinations enabled in production. To verify implementation
      without feature flag enabled it should be disabled explicitly in the
      test.
      
      Part of: gitlab-org/gitaly#2702
      a1731804
  4. 05 Aug, 2020 22 commits