# Ziti ha cluster - sdk enabled apps falling off **URL:** \ **Category:** Ziti Overlay\ **Created:** [July 25, 2025, 2:00am UTC](https://openziti.discourse.group/t/ziti-ha-cluster-sdk-enabled-apps-falling-off/4900 "2025-07-25T02:00:50Z")\ **Posts on this page:** 6\ **Page:** 1
**Author:** ![strongwazz](https://yyz2.discourse-cdn.com/free1/user_avatar/openziti.discourse.group/strongwazz/32/3630_2.png) [@strongwazz](https://openziti.discourse.group/u/strongwazz)\ **Post date:** [July 25, 2025, 2:00am UTC](https://openziti.discourse.group/t/ziti-ha-cluster-sdk-enabled-apps-falling-off/4900/1 "2025-07-25T02:00:50Z")
Hiya I have a 3-node HA controller cluster, with 3 edge routers This works brilliantly when working with tunnelers (ZDE etc) When using SDK-enabled/aware apps that host a service, they eventually fall off. For testing, if I run 1 controller (in HA), the app works fine. The issue only occurs when I introduce two or more controllers into the cluster. The app complains about connecting to the edge routers; however, between the two tests (one HA controller vs. two or more HA controllers), nothing changes regarding the edge routers. I've tested running my "01" controller in an HA configuration as a single controller, as well as "02", so I know base network communication between controllers and edge routers is ok. The controllers are all within the same LAN/VLAN, with no host-level firewalls, so communication between them is direct and unrestricted. I've tried using binaries 1.5.4 as well as 1.6.5, with no material difference. Currently using go-sdk 1.2.0, but did also try a previous 1.1.x version ```auto INFO[0000] new service session session token=642a526e-fb52-xxxx-xxxx-d201755e8457 ERRO[0056] unable to unbind session for conn connId=1 error="channel closed" sessionId=642a526e-fb52-xxxx-xxxx-d201755e8457 ERRO[0056] failed to close listener connId=1 error="channel closed" marker= serviceName=web-service ERRO[0059] unable to unbind session for conn connId=1 error="timeout waiting for message to be written to wire: context deadline exceeded" sessionId=642a526e-fb52-xxxx-xxxx-d201755e8457 ERRO[0059] failed to close listener connId=1 error="timeout waiting for message to be written to wire: context deadline exceeded" marker= serviceName=web-service ERRO[0059] failed to bind _context="ch{ziti-sdk[router=tls:01.edge.domain:3022]}->u{classic}->i{m6AYZQI4SJ/QjJQ}" connId=1 error="timeout waiting for message reply: context deadline exceeded" serviceName=web-service sessionId=cmdi56gwj019wscn3orgevjsx ERRO[0059] failed to establish listener connId=1 error="timeout waiting for message reply: context deadline exceeded" router=01.edge.domain serviceId=qp9K4gnL2rrewFIUMAEgw serviceName=web-service ERRO[0059] creating listener failed after 5001ms: timeout waiting for message reply: context deadline exceeded router=01.edge.domain serviceName=web-service INFO[0059] notify error handler of error: timeout waiting for message reply: context deadline exceeded ``` I want to emphasise that the edge router configuration remains unchanged between tests. Only when multiple controllers are in play do the issues occur. And it only affects SDK-based apps, tunnelling identities with ZDE or similar works flawlessly in both cases. I'd be happy to share configs and run any other diagnostics with developers if it's deemed useful. I just want to do that in a slightly less public fashion due to commercials. Or, what glaringly obvious thing am I overlooking? =D tia \<3 ---
**Author:** ![plorenz](https://yyz2.discourse-cdn.com/free1/user_avatar/openziti.discourse.group/plorenz/32/54_2.png) [@plorenz](https://openziti.discourse.group/u/plorenz)\ **Post date:** [July 25, 2025, 3:39am UTC](https://openziti.discourse.group/t/ziti-ha-cluster-sdk-enabled-apps-falling-off/4900/2 "2025-07-25T03:39:45Z")
Hi @strongwazz The Go SDK still requires that the Enable HA flag is set, see: [sdk-golang/ziti/config.go at v1.2.0 路 openziti/sdk-golang 路 GitHub](https://github.com/openziti/sdk-golang/blob/v1.2.0/ziti/config.go#L51) We can hopefully soon make that flag obsolete. Let us know if that doesn't fix things and/or if you already had that flag set 馃檪 Paul ---
**Author:** ![strongwazz](https://yyz2.discourse-cdn.com/free1/user_avatar/openziti.discourse.group/strongwazz/32/3630_2.png) [@strongwazz](https://openziti.discourse.group/u/strongwazz)\ **Post date:** [July 25, 2025, 6:18am UTC](https://openziti.discourse.group/t/ziti-ha-cluster-sdk-enabled-apps-falling-off/4900/3 "2025-07-25T06:18:46Z")
Haha, such a wonderful oversight on my part. Thanks, ill give that a crack ---
**Author:** ![strongwazz](https://yyz2.discourse-cdn.com/free1/user_avatar/openziti.discourse.group/strongwazz/32/3630_2.png) [@strongwazz](https://openziti.discourse.group/u/strongwazz)\ **Post date:** [July 27, 2025, 11:40pm UTC](https://openziti.discourse.group/t/ziti-ha-cluster-sdk-enabled-apps-falling-off/4900/4 "2025-07-27T23:40:42Z")
Nailed it 馃檪 Didnt realise it was a flag that needed setting ---
**Author:** ![plorenz](https://yyz2.discourse-cdn.com/free1/user_avatar/openziti.discourse.group/plorenz/32/54_2.png) [@plorenz](https://openziti.discourse.group/u/plorenz)\ **Post date:** [July 28, 2025, 1:46pm UTC](https://openziti.discourse.group/t/ziti-ha-cluster-sdk-enabled-apps-falling-off/4900/5 "2025-07-28T13:46:45Z")
Excellent 馃檪 I added an issue to make sure we don鈥檛 lose track of removing the need for the EnableHA flag: [Remove need to EnableHA flag in Go SDK 路 Issue #779 路 openziti/sdk-golang 路 GitHub](https://github.com/openziti/sdk-golang/issues/779) ---
**Author:** ![vladsh](https://avatars.discourse-cdn.com/v4/letter/v/f04885/32.png) [@vladsh](https://openziti.discourse.group/u/vladsh)\ **Post date:** [August 15, 2025, 9:39am UTC](https://openziti.discourse.group/t/ziti-ha-cluster-sdk-enabled-apps-falling-off/4900/6 "2025-08-15T09:39:46Z")
Hey @strongwazz and Ziti maintainers\*\* I鈥檓 building pretty similar setup: 3 VMs with Controller and Routers on it These Controllers/Routers VMs are in different regions: EU, Asia, and US and a set of other routers, tunellers, sidecars from different k8s clusters connecting to the controllers. The issue I found is the speed of connections. Before deploying this HA setup, I did a PoC with a single Controller and it feels to me like everything was way faster. Could you describe how far/close your controllers to each other? Also any recommendations you have learn much appreciated) Regards \*\*