make success_report_interval configuraable

[hippotat.git] / README.config

diff --git a/README.config b/README.config
index cde9e130d248f5e183338d3e76960ca4839e16f8..aa493bb7a41db77108259cb2c167bb03a00cf661 100644 (file)
--- a/README.config
+++ b/README.config
@@ -71,11 +71,13 @@ Exceptional settings:
      Command to run to create and communicate with local network
      interface.  Passed to sh -c.  Must speak SLIP on stdin/stdout.
      The following interpolations aare substituted:
-                       %(local)s  %(peer)s  %(rnets)s   %(ifname)s
+                       %{local}   %{peer}   %{rnets}    %{ifname}
           on server    <vaddr>    <vrelay>  <vnetwork>  <ifname_server>
           on client    <client>   <vaddr>   <vroutes>   <ifname_client>
-     Plus %(mtu)s (and %% to indicate a literal %).
-     ["userv root ipif %(local)s,%(peer)s,%(mtu)s,slip '%(rnets)s'"]
+     Plus %{mtu} and %% to indicate a literal %.
+     (For compatibility with older hippotat, %(var)s is supported too
+     but this is deprecated since the extra `s` is confusing.)
+     ["userv root ipif %{local},%{peer},%{mtu},slip '%{rnets}'"]
 
      On server: applies to all clients; not looked up in
       client-specific sections.
@@ -87,7 +89,11 @@ Capped settings:
      those from the other sections (including COMMON).
 
   max_batch_down
-     Size limit for response payloads (used by server only)
+     Size limit for response payloads.
+     On client, incoming response bodies are limited to this plus
+     a fixed constant metadata overhead of 10000 bytes.
+     Server uses minim of client and server value (old servers
+     just uses server's value).
      [65536 bytes; LIMIT: 262144 bytes]
 
   max_queue_time
@@ -104,8 +110,10 @@ Capped settings:
       for this long
      On client: give up on any http request outstanding for
       for this long plus http_timeout_grace
+     Warning messages about link problems, printed by the client,
+     are rate limited to no more than one per effective timeout.
      Client's effective timeout must be at least server's (checked).
-     [30 s; LIMIT: 121]
+     [30 s; LIMIT: 121s]
 
   target_requests_outstanding   
      On server: whenever number of outstanding requests for
@@ -176,6 +184,10 @@ Ordinary settings, used by client only:
   max_batch_up
      Size limit for request payloads. [4000 bytes]
 
+  success_report_interval
+     If nonzero, report success periodically.  Otherwise just
+     report it when we first have success.  [3600 s]
+
   http_retry
      If a request fails, wait this long before considering it
      "finished" - to limit rate of futile requests.  [5 s]