# How should GCache size be determined for PXC?

**URL:** <https://forums.percona.com/t/how-should-gcache-size-be-determined-for-pxc/41230>\
**Category:** Percona XtraDB Cluster 8.x\
**Tags:** mysql, percona\
**Created:** [September 2, 2026, 10:23am UTC](https://forums.percona.com/t/how-should-gcache-size-be-determined-for-pxc/41230 "2026-09-02T10:23:25Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![Sekhri](https://sea1.discourse-cdn.com/flex019/user_avatar/forums.percona.com/sekhri/32/24194_2.png) [@Sekhri](https://forums.percona.com/u/Sekhri)\
**Post date:** [September 2, 2026, 10:23am UTC](https://forums.percona.com/t/how-should-gcache-size-be-determined-for-pxc/41230/1 "2026-09-02T10:23:25Z")

</div>

I am running a 3-node Percona XtraDB Cluster on Kubernetes, where the PV size is selected dynamically by the user on the dashboard (minimum 50GB).

What is the recommended approach for determining `gcache.size`?

Is there a recommended formula or guideline for calculating `gcache.size` when the workload is not known in advance?

---

<div class="post-metadata">

**Author:** ![Wayne\_Leutwyler](https://sea1.discourse-cdn.com/flex019/user_avatar/forums.percona.com/wayne_leutwyler/32/20829_2.png) [@Wayne\_Leutwyler](https://forums.percona.com/u/Wayne_Leutwyler)\
**Post date:** [September 2, 2026, 11:22am UTC](https://forums.percona.com/t/how-should-gcache-size-be-determined-for-pxc/41230/2 "2026-09-02T11:22:20Z")

</div>

This blog post: [https://www.percona.com/blog/calculate-correct-size-percona-xtradb-clusters-gcache/](https://www.percona.com/blog/calculate-correct-size-percona-xtradb-clusters-gcache/) should provide you the information you are looking for.

---

<div class="post-metadata">

**Author:** ![dogs\_Cute](https://sea1.discourse-cdn.com/flex019/user_avatar/forums.percona.com/dogs_cute/32/12754_2.png) [@dogs\_Cute](https://forums.percona.com/u/dogs_Cute)\
**Post date:** [September 2, 2026, 1:10pm UTC](https://forums.percona.com/t/how-should-gcache-size-be-determined-for-pxc/41230/3 "2026-09-02T13:10:14Z")

</div>

Detailed calculation method:

> **[Tuning gcache.size - Rapid Node Recovery with IST and the GCache | Galera...](https://mariadb.com/docs/galera-cluster/high-availability/rapid-node-recovery-with-ist-and-the-gcache#tuning-gcache.size)**
>
> This page provides a deep-dive into Incremental State Transfer (IST), a method for a node to synchronize with the cluster. For information on a fallback mechanism, see State Snapshot Transfers (SSTs). | Incremental State Transfer (IST) recovers a...

However, because the cluster needs to be upgraded, in general, it is set to 2G.

---

<div class="post-metadata">

**Author:** ![przemek](https://sea1.discourse-cdn.com/flex019/user_avatar/forums.percona.com/przemek/32/3_2.png) [@przemek](https://forums.percona.com/u/przemek)\
**Post date:** [September 7, 2026, 7:46am UTC](https://forums.percona.com/t/how-should-gcache-size-be-determined-for-pxc/41230/4 "2026-09-07T07:46:27Z")

</div>

And if you are interested later in what’s fitting exactly in the Gcache, this may be handy: [GitHub - PrzemekMalkowski/gcache-inspector: Offline reader and decoder for Galera write-set cache (galera.cache) · GitHub](https://github.com/PrzemekMalkowski/gcache-inspector)