From a01b1f036d4b2218b79e2e9134bf9a12e004cd92 Mon Sep 17 00:00:00 2001
From: Francesco Mazzoli <francesco.mazzoli@xtxmarkets.com>
Date: Thu, 27 Jul 2023 13:54:51 +0000
Subject: [PATCH] More alert-related fixes

---
 cpp/cdc/CDC.cpp       | 2 +-
 cpp/shard/Shard.cpp   | 7 ++++++-
 go/lib/log.go         | 2 +-
 go/lib/metadatareq.go | 4 +++-
 4 files changed, 11 insertions(+), 4 deletions(-)

diff --git a/cpp/cdc/CDC.cpp b/cpp/cdc/CDC.cpp
index 5c786c5a..7654976e 100644
--- a/cpp/cdc/CDC.cpp
+++ b/cpp/cdc/CDC.cpp
@@ -540,7 +540,7 @@ private:
             int err = errno;
             // Note that we get EPERM on `sendto` when nf drops packets.
             if (err == EAGAIN || err == EPERM) {
-                _env.raiseAlert(alert, false, "we got %s/%s=%s when trying to send shard message, will wait and retry", err, translateErrno(err), safe_strerror(err));
+                alert = _env.raiseAlert(alert, false, "we got %s/%s=%s when trying to send shard message, will wait and retry", err, translateErrno(err), safe_strerror(err));
                 sleepFor(100_ms);
             } else {
                 throw EXPLICIT_SYSCALL_EXCEPTION(err, "sendto");
diff --git a/cpp/shard/Shard.cpp b/cpp/shard/Shard.cpp
index 347c424d..1d0fd3a9 100644
--- a/cpp/shard/Shard.cpp
+++ b/cpp/shard/Shard.cpp
@@ -307,7 +307,12 @@ public:
             }
 
             if (sendto(sock, respBbuf.data, respBbuf.len(), 0, (struct sockaddr*)&clientAddr, sizeof(clientAddr)) != respBbuf.len()) {
-                throw SYSCALL_EXCEPTION("sendto");
+                // we get this when nf drops packets
+                if (errno != EPERM) {
+                    throw SYSCALL_EXCEPTION("sendto");
+                } else {
+                    LOG_INFO(_env, "dropping response %s to %s because of EPERM", respContainer->kind(), clientAddr);
+                }
             }
             LOG_DEBUG(_env, "sent response %s to %s", respContainer->kind(), clientAddr);
         }
diff --git a/go/lib/log.go b/go/lib/log.go
index 3fd1c4f5..aa1a1f7b 100644
--- a/go/lib/log.go
+++ b/go/lib/log.go
@@ -361,7 +361,7 @@ func (l *Logger) NewNCAlert() *NCAlert {
 }
 
 func (nc *NCAlert) Alert(f string, v ...any) {
-	if nc.alert == nil {
+	if nc.alert == nil && nc.l.troll != nil {
 		nc.alert = nc.l.troll.NewUnbinnableAlertStatus()
 	}
 	nc.l.LogStack(1, ERROR, "nc alert "+f, v...)
diff --git a/go/lib/metadatareq.go b/go/lib/metadatareq.go
index ad1d56a3..37f2d0af 100644
--- a/go/lib/metadatareq.go
+++ b/go/lib/metadatareq.go
@@ -170,7 +170,7 @@ func (c *Client) metadataRequestInternal(
 				if dontWait {
 					log.Info("dontWait is on, we couldn't send the request due to EPERM %v, goodbye", err)
 				} else {
-					epermAlert.Alert("got possibly transient EPERM when sending to shard %v, might retry: %v", shid, err)
+					epermAlert.Alert("got possibly transient EPERM when sending to shard %v, might retry after waiting for %v: %v", shid, timeout, err)
 					time.Sleep(timeout)
 					attempts++
 					continue
@@ -178,6 +178,8 @@ func (c *Client) metadataRequestInternal(
 			} else {
 				return fmt.Errorf("couldn't send request to shard %v: %w", shid, err)
 			}
+		} else {
+			epermAlert.Clear()
 		}
 		if written < len(reqBytes) {
 			panic(fmt.Sprintf("incomplete send to shard %v -- %v bytes written instead of %v", shid, written, len(reqBytes)))