Hello, we get this error every once in a while on one of our servers. This failure then results in the notify sending host to not try again. Would it not be better to use NSD_RC_DISCARD in this place so the server can try again or am I too naive? Beside the fact that this happening may be a bug (I have failed so far finding the root cause), it doesn't appear to be fatal for nsd, however, since a notify-reply with SERVFAIL is sent, it is fatal for that specific domain. Alf static query_state_type answer_notify(struct nsd* nsd, struct query *query) { ... if(!write_socket(s, &mode, sizeof(mode)) || !write_socket(s, &sz, sizeof(sz)) || !write_socket(s, buffer_begin(query->packet), buffer_limit(query->packet)) || !write_socket(s, &acl_send, sizeof(acl_send)) || !write_socket(s, &acl_xfr, sizeof(acl_xfr))) { log_msg(LOG_ERR, "error in IPC notify server2main, %s", strerror(errno)); return query_error(query, NSD_RC_SERVFAIL); } ...