Temporary fix: Avoid killing Coordinator with unregistered cache mgr actions

that cause isOpen() assertions.

If a worker forwards a cache manager request to Coordinator and Coordinator
does not have that action registered, CacheManager::createRequestedAction()
throws (as it should) and Mgr::Request cleanup asserts when its half-baked
connection tries to close a not-yet-imported socket descriptor.

This workaround catches the exception, reports it, and manually closes the
socket descriptor. It also prevents an ACK response from being sent to the
worker, which triggers a worker timeout.

Mid-term TODO: Coordinator should register all actions that are known to kids.
Should Coordinator respond with an error instead of relying on a timeout?

Long-term TODO: Consider an API where cache manager responses can be
aggregated and formatted by Coordinator without knowing action-specific
details. After all, there are not so many types of action information (size,
count, rate, etc.) and most actions have simple reporting formats. Currently,
it is awkward to guarantee that Coordinator and all workers know all actions,
especially when some actions may be specific to non-worker kids such as
Coordinator and diskers.

diff --git a/src/ipc/Coordinator.cc b/src/ipc/Coordinator.cc
index 6379f4b59e42048ebf3411d8123269ee8c8d4a16..7bc106ffce2c26a4a6abcedbf88f5d840112ae34 100644 (file)
--- a/src/ipc/Coordinator.cc
+++ b/src/ipc/Coordinator.cc
@@ -164,15 +164,26 @@ Ipc::Coordinator::handleCacheMgrRequest(const Mgr::Request& request)
 {
     debugs(54, 4, HERE);
 
+    try {
+        Mgr::Action::Pointer action =
+            CacheManager::GetInstance()->createRequestedAction(request.params);
+        AsyncJob::Start(new Mgr::Inquirer(action, request, strands_));
+    }
+    catch (const std::exception &ex) {
+        debugs(54, DBG_IMPORTANT, "BUG: cannot aggregate mgr:" <<
+               request.params.actionName << ": " << ex.what());
+        // TODO: Avoid half-baked Connections or teach them how to close.
+        ::close(request.conn->fd);
+        request.conn->fd = -1;
+        return; // the worker will timeout and close
+    }
+
     // Let the strand know that we are now responsible for handling the request
     Mgr::Response response(request.requestId);
     TypedMsgHdr message;
     response.pack(message);
     SendMessage(MakeAddr(strandAddrPfx, request.requestorId), message);
 
-    Mgr::Action::Pointer action =
-        CacheManager::GetInstance()->createRequestedAction(request.params);
-    AsyncJob::Start(new Mgr::Inquirer(action, request, strands_));
 }
 
 void