Fix haproxy non-agent healthchecks to test something that requires both the

postgrest-rewriter and the postgrest instance to be working.  Our previous
checks just tested whether the rewriter was working, and it could still
succeed even when the postgrest instance it was proxying was down.

haproxy/haproxy.cfg

@@ -82,14 +82,29 @@ backend hived-http
 
 backend balance-tracker
   balance roundrobin
+  # balance tracker doesn't have a cheap API method we can call, so test whether we can get
+  # the OpenAPI JSON
+  option httpchk
+  http-check connect
+  http-check send meth HEAD uri /
+  http-check expect status 200
+  timeout check 100
   server balance-tracker balance-tracker-postgrest-rewriter:80 check agent-check agent-addr haproxy-healthchecks agent-port 7004 agent-inter 10s
 
 backend reputation-tracker
   balance roundrobin
+  option httpchk
+  http-check connect
+  http-check send meth GET uri /last-synced-block
+  http-check expect status 200
   server reputation-tracker reputation-tracker-postgrest-rewriter:80 check agent-check agent-addr haproxy-healthchecks agent-port 7009 agent-inter 10s
 
 backend hafah
   balance roundrobin
+  option httpchk
+  http-check connect
+  http-check send meth GET uri /version
+  http-check expect status 200
   server hafah-postgrest hafah-postgrest-rewriter:80 check agent-check agent-addr haproxy-healthchecks agent-port 7003 agent-inter 10s
 
 backend hivemind
@@ -98,6 +113,10 @@ backend hivemind
 
 backend block-explorer
   balance roundrobin
+  option httpchk
+  http-check connect
+  http-check send meth GET uri /last-synced-block
+  http-check expect status 200
   server block-explorer block-explorer-postgrest-rewriter:80 check agent-check agent-addr haproxy-healthchecks agent-port 7005 agent-inter 10s
 
 backend sync-hived