Shared archives between replica and master

Typical replication Shared archives

Typically, a replicaClosed A Helix Core Server that automatically maintains a full or partial copy of metadata and none, some, or all of the related file archives by copying data from a master Helix Core Server using 'p4 pull' or 'p4 journalcopy'. server retrieves both its metadata and file archives from the master server on the user-defined pull interval. For example p4 pull -i 1

This configuration requires the P4TARGET server to send the archives files to the replica.

If a replica server is configured to share the same physical archive files as the master server:

  • the replica accesses the archives directly, so archive files are not transferred
  • the replica and master can:
    • run on the same machine, or
    • share storage over a network shared storage, where network latency affects performance.
Note

Sharing archives is supported between a master and replica server, or between a commit server and edge server.

Replica servers can't share archives with other replica servers.

To share archives, on the replica, set the lbr.replication configurable to shared either in the server spec or manually:

  • only metadata is retrieved on the pull interval
  • the "shared" archive files are not retrieved until requested by a client
  • new files are not automatically transferred
  • purged files are not removed

Shared archives can form part of a High Availability configuration.

Warning

When archive files are directly shared between a replica and master server, the replica must have lbr.replication set to shared. Otherwise, the files in the archive might be corrupted.

To configure a replica to share archive files with a master

  1. Ensure that the clocks for the master and replica servers are synchronized.

    Nothing needs to be done if the master and replica servers are hosted on the same operating system.

    Synchronizing clocks is a system administration task that typically involves using a Network Time Protocol client to synchronize an operating system’s clock with a time server on the Internet, or a time server you maintain for your own network.

    See http://support.ntp.org/bin/view/Support/InstallingNTP.

  2. If you have not already done so, configure the replica or edge server. See Deployment architecture.

  3. Set server.depot.root on the master and replica, or the commit and edge, to point to the shared archive location for relative depot map paths to use. For example,

    p4 configure set master#server.depot.root=/data/depots
    p4 configure set replica#server.depot.root=/data/depots

    Note

    Review any absolute depot map paths to ensure the replica server has access to write files to that absolute shared archive location. The replica server enforces correct lbr.replication settings when archives are being shared. Therefore, the replica server needs to write files to the shared archive locations noted by any depot map paths. See Providing map information in Helix Core Command-Line (P4) Reference.

  4. Set lbr.replication=shared either in the replica's server spec or manually using a command similar to this:

    p4 configure set replica#lbr.replication=shared

  5. Restart the replica.

Result

The result of this configuration:

  • archive file content is only retrieved when requested, and those requests are made against the shared archives
  • commands that would schedule the transfer of file content, such as p4 pull -u and p4 verify -t are rejected

  • if startup configurables, such as startup.N=pull -u, are defined, the replica server attempts to run such commands. Because the attempt to retrieve archive content is rejected, the replica’s server log will contain an error:

    Perforce server error:
            2014/01/23 13:02:31 pid 6952 service-od@21131 background 'pull -u -i 10'
    Note

    For upgrading, see Upgrading replica servers.