staging garage: use new health check endpoint

1 files changed, 24 insertions, 18 deletions

diff --git a/cluster/staging/app/garage/deploy/garage-nix2.hcl b/cluster/staging/app/garage/deploy/garage-nix2.hcl
index a137881..7ef4346 100644
--- a/cluster/staging/app/garage/deploy/garage-nix2.hcl
+++ b/cluster/staging/app/garage/deploy/garage-nix2.hcl
@@ -25,7 +25,7 @@ job "garage-staging" {
       config {
         packages = [
           "#bash", # so that we can enter a shell inside container
-          "git+https://git.deuxfleurs.fr/Deuxfleurs/garage.git?ref=main&rev=19bdd1c7992f907c9bd518462054f56e28a2e45b",
+          "git+https://git.deuxfleurs.fr/Deuxfleurs/garage.git?ref=admin-health-api&rev=d7868c48a4d8d5831051a0be088fe7bbec259bca",
         ]
         command = "garage"
         args = [ "server" ]
@@ -75,8 +75,6 @@ job "garage-staging" {
         name = "garage-staging-rpc"
         tags = ["garage-staging-rpc"]
         port = "rpc"
-        # No check on RPC, it wouldn't try connecting to the correct address
-        # (Garage listens only on IPv6 for RPC, see config file)
       }
 
       service {
@@ -88,13 +86,11 @@ job "garage-staging" {
         ]
         port = "s3"
         check {
-          type = "tcp"
+          port = "admin"
+          type = "http"
+          path = "/health"
           interval = "60s"
           timeout = "5s"
-          check_restart {
-            limit = 3
-            grace = "90s"
-          }
         }
       }
 
@@ -107,13 +103,11 @@ job "garage-staging" {
         ]
         port = "k2v"
         check {
-          type = "tcp"
+          port = "admin"
+          type = "http"
+          path = "/health"
           interval = "60s"
           timeout = "5s"
-          check_restart {
-            limit = 3
-            grace = "90s"
-          }
         }
       }
 
@@ -128,13 +122,11 @@ job "garage-staging" {
         ]
         port = "web"
         check {
-          type = "tcp"
+          port = "admin"
+          type = "http"
+          path = "/health"
           interval = "60s"
           timeout = "5s"
-          check_restart {
-            limit = 3
-            grace = "90s"
-          }
         }
       }
 
@@ -145,12 +137,26 @@ job "garage-staging" {
         ]
         port = "admin"
         check {
+          name = "garage-admin-health-check"
+          type = "http"
+          path = "/health"
+          interval = "60s"
+          timeout = "5s"
+          check_restart {
+            limit = 10
+            grace = "90s"
+            ignore_warnings = true
+          }
+        }
+        check {
+          name = "garage-tcp-liveness-check"
           type = "tcp"
           interval = "60s"
           timeout = "5s"
           check_restart {
             limit = 3
             grace = "90s"
+            ignore_warnings = true
           }
         }
       }