We put ProxySQL between your application and MySQL or MariaDB so connection storms, replica lag and a primary failover stop being application problems. Pooling, read/write splitting and routing rules, set up and tested against your real traffic.
Pooled, routed and failover-aware, with the rules written down.
Thousands of application connections multiplexed onto a small, steady set of backend connections, so MySQL spends its time on queries.
Query rules send writes to the primary and safe reads to replicas, with no changes to application code.
Hostgroups that follow replication, Galera or Group Replication topology, so a promoted primary takes traffic without a config push.
Replicas that fall behind are taken out of rotation automatically, so users do not read stale data.
Cache hot read queries at the proxy, and rewrite or block a bad query pattern without waiting on a deploy.
ProxySQL's own stats show which query digests cost the most, feeding monitoring and the next round of tuning.
We run ProxySQL in front of our own MariaDB cluster. The configuration we hand over is the kind we have had to rely on ourselves.
Usually only the database host and port. ProxySQL speaks the MySQL protocol, so the app connects to it as if it were the database, and routing is decided by rules on the proxy.
It depends on how your app reads its own writes. We start with everything on the primary, then move specific read queries to replicas by rule, checking each for stale-read risk rather than splitting everything at once.
ProxySQL watches the backends and moves the writer role to the promoted server. Promotion itself is done by your replication manager or cluster; we wire the two together and rehearse the failover so the behaviour is known.
Only if it is deployed as one. We run it either alongside each application instance or as a redundant pair, so losing a proxy does not take the database path with it.
Tell us what you're running — we'll tell you where the wins are before you commit.
Architecture → implementation → proof → runbook · senior US engineers · $1M insured
Emergency support is our entry tier — existing client or not, start an urgent request and a senior engineer will jump on it.