Tuning GlusterFS geo-replication session options keeps asynchronous replication aligned with recovery goals and link capacity. Settings such as concurrency and delete handling influence how quickly the secondary catches up and how much bandwidth is consumed during steady state and catch-up syncs.
Geo-replication tracks changes on the primary and ships them to a paired secondary using gsyncd workers and a configured transfer method (for example rsync or tar-over-SSH). Session settings are stored per primary/secondary pair, and gluster volume geo-replication … config is used to list option/value pairs or update a single option for that session.
Some option changes can restart the geo-replication session and trigger additional crawling, which may temporarily increase load on bricks and network links. Options such as ignore-deletes change retention semantics on the secondary, so operational expectations for restores and cleanup should be confirmed before enabling them.
$ sudo gluster volume geo-replication gvol-primary geoaccount@snode1.example.com::gvol-secondary config sync-jobs 3 log-level INFO ignore-deletes 0
Option names and defaults can vary by release. Copy the exact key name from the listing when updating values.
$ sudo gluster volume geo-replication gvol-primary geoaccount@snode1.example.com::gvol-secondary config sync-jobs 6 geo-replication config updated successfully
sync-jobs increases concurrency per active worker and can raise CPU, disk, and WAN utilization when set too high.
$ sudo gluster volume geo-replication gvol-primary geoaccount@snode1.example.com::gvol-secondary config ignore-deletes 1 geo-replication config updated successfully
Enabling ignore-deletes makes the secondary a superset of the primary and can complicate retention and cleanup.
$ sudo gluster volume geo-replication gvol-primary geoaccount@snode1.example.com::gvol-secondary config sync-jobs 6 log-level INFO ignore-deletes 1