# Why can’t the haproxy-replicas Service be used for write requests?

**URL:** <https://forums.percona.com/t/why-can-t-the-haproxy-replicas-service-be-used-for-write-requests/25061>\
**Category:** Percona Operator for MySQL\
**Created:** [September 10, 2023, 1:37am UTC](https://forums.percona.com/t/why-can-t-the-haproxy-replicas-service-be-used-for-write-requests/25061 "2023-09-10T01:37:25Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![gmautner](https://avatars.discourse-cdn.com/v4/letter/g/71c47a/32.png) [@gmautner](https://forums.percona.com/u/gmautner)\
**Post date:** [September 10, 2023, 1:37am UTC](https://forums.percona.com/t/why-can-t-the-haproxy-replicas-service-be-used-for-write-requests/25061/1 "2023-09-10T01:37:25Z")

</div>

In the documentation from [Custom Resource options - Percona Operator for MySQL based on Percona XtraDB Cluster](https://docs.percona.com/percona-operator-for-mysql/pxc/operator.html)

You state that the haproxy-replicas should not be used for write requests.

Where does that come from? Doesn’t the XtraDB cluster use a multi-master Galera cluster which enables writing to all nodes?

If for some reason it’s not recommended, is it enforceable? Because once you offer the additional service to developers, you can’t just count on people to respect that restriction.

---

<div class="post-metadata">

**Author:** ![Slava\_Sarzhan](https://sea1.discourse-cdn.com/flex019/user_avatar/forums.percona.com/slava_sarzhan/32/3196_2.png) [@Slava\_Sarzhan](https://forums.percona.com/u/Slava_Sarzhan)\
**Post date:** [September 11, 2023, 6:05pm UTC](https://forums.percona.com/t/why-can-t-the-haproxy-replicas-service-be-used-for-write-requests/25061/2 "2023-09-11T18:05:44Z")

</div>

@gmautner As you know XtraDB cluster uses a multi-master Galera cluster which enables writing to all nodes and you can use haproxy-replicas for writes. We did not add any limitations to it. But as a result, the performance of your cluster will be not good. This service was created to have the possibility to read from all available PXC pods. Using `cluster_name-haproxy` service you can’t get this functionality.