Fix error fingerprinting (#2237)

Author: chadwhitacre

_unit-test/error-handling-test.sh

@@ -14,48 +14,73 @@ send_envelope() {
   echo "Test Sending $1"
 }
 
+##########################
+
 export -f send_envelope
 echo "Testing initial send_event"
 export log_file=test_log.txt
 echo "Test Logs" >"$log_file"
 echo "Error Msg" >>"$log_file"
 breadcrumbs=$(generate_breadcrumb_json | sed '$d' | $jq -s -c)
-SEND_EVENT_RESPONSE=$(send_event "12345123451234512345123451234512" "Test exited with status 1" "{\"ignore\": \"me\"}" "$breadcrumbs")
+SEND_EVENT_RESPONSE=$(
+  send_event \
+    "'foo' exited with status 1" \
+    "Test exited with status 1" \
+    "Traceback: ignore me" \
+    "{\"ignore\": \"me\"}" \
+    "$breadcrumbs"
+)
 rm "$log_file"
-test "$SEND_EVENT_RESPONSE" == 'Test Sending sentry-envelope-12345123451234512345123451234512'
-ENVELOPE_CONTENTS=$(cat /tmp/sentry-envelope-12345123451234512345123451234512)
-test "$ENVELOPE_CONTENTS" == "$(cat _unit-test/snapshots/sentry-envelope-12345123451234512345123451234512)"
+expected_filename='sentry-envelope-f73e4da437c42a1d28b86a81ebcff35d'
+test "$SEND_EVENT_RESPONSE" == "Test Sending $expected_filename"
+ENVELOPE_CONTENTS=$(cat "/tmp/$expected_filename")
+test "$ENVELOPE_CONTENTS" == "$(cat _unit-test/snapshots/$expected_filename)"
 echo "Pass."
 
+##########################
+
 echo "Testing send_event duplicate"
-SEND_EVENT_RESPONSE=$(send_event "12345123451234512345123451234512" "Test exited with status 1" "{\"ignore\": \"me\"}" "$breadcrumbs")
+SEND_EVENT_RESPONSE=$(
+  send_event \
+    "'foo' exited with status 1" \
+    "Test exited with status 1" \
+    "Traceback: ignore me" \
+    "{\"ignore\": \"me\"}" \
+    "$breadcrumbs"
+)
 test "$SEND_EVENT_RESPONSE" == "Looks like you've already sent this error to us, we're on it :)"
 echo "Pass."
-rm '/tmp/sentry-envelope-12345123451234512345123451234512'
+rm "/tmp/$expected_filename"
+
+##########################
 
 echo "Testing cleanup without minimizing downtime"
 export REPORT_SELF_HOSTED_ISSUES=0
 export MINIMIZE_DOWNTIME=''
 export dc=':'
 echo "Test Logs" >"$log_file"
-CLEANUP_RESPONSE=$(cleanup ERROR)
+CLEANUP_RESPONSE=$(cleanup ERROR) # the linenumber of this line must match just below
 rm "$log_file"
-test "$CLEANUP_RESPONSE" == 'Error in _unit-test/error-handling-test.sh:41.
+test "$CLEANUP_RESPONSE" == 'Error in _unit-test/error-handling-test.sh:62.
 '\''local cmd="${BASH_COMMAND}"'\'' exited with status 0
 
 Cleaning up...'
 echo "Pass."
 
+##########################
+
 echo "Testing cleanup while minimizing downtime"
 export REPORT_SELF_HOSTED_ISSUES=0
 export MINIMIZE_DOWNTIME=1
 echo "Test Logs" >"$log_file"
-CLEANUP_RESPONSE=$(cleanup ERROR)
+CLEANUP_RESPONSE=$(cleanup ERROR) # the linenumber of this line must match just below
 rm "$log_file"
-test "$CLEANUP_RESPONSE" == 'Error in _unit-test/error-handling-test.sh:53.
+test "$CLEANUP_RESPONSE" == 'Error in _unit-test/error-handling-test.sh:76.
 '\''local cmd="${BASH_COMMAND}"'\'' exited with status 0
 
 *NOT* cleaning up, to clean your environment run "docker compose stop".'
 echo "Pass."
 
+##########################
+
 report_success

_unit-test/snapshots/sentry-envelope-12345123451234512345123451234512 renamed to _unit-test/snapshots/sentry-envelope-f73e4da437c42a1d28b86a81ebcff35d

@@ -1,6 +1,6 @@
-{"event_id":"12345123451234512345123451234512","dsn":"https://[email protected]/3"}
+{"event_id":"f73e4da437c42a1d28b86a81ebcff35d","dsn":"https://[email protected]/3"}
 {"type":"event"}
-{"level":"error","exception":{"values":[{"type":"Error","value":"Test exited with status 1","stacktrace":{"frames":[{"ignore":"me"}]}}]},"breadcrumbs":{"values":[{"message":"Test Logs","category":"log","level":"info"}]}}
+{"level":"error","exception":{"values":[{"type":"Error","value":"Test exited with status 1","stacktrace":{"frames":[{"ignore":"me"}]}}]},"breadcrumbs":{"values":[{"message":"Test Logs","category":"log","level":"info"}]},"fingerprint":["f73e4da437c42a1d28b86a81ebcff35d"]}
 {"type":"attachment","length":20,"content_type":"text/plain","filename":"install_log.txt"}
 Test Logs
 Error Msg

install/error-handling.sh

@@ -20,11 +20,17 @@ generate_breadcrumb_json() {
 
 send_event() {
   # Use traceback hash as the UUID since it is 32 characters long
-  local event_hash=$1
+  local cmd_exit=$1
   local error_msg=$2
-  local traceback_json=$3
-  local breadcrumbs=$4
-  local envelope_file="sentry-envelope-${event_hash}"
+  local traceback=$3
+  local traceback_json=$4
+  local breadcrumbs=$5
+  local fingerprint_value=$(
+    echo -n "$cmd_exit $error_msg $traceback" |
+      docker run -i --rm busybox md5sum |
+      cut -d' ' -f1
+  )
+  local envelope_file="sentry-envelope-${fingerprint_value}"
   local envelope_file_path="/tmp/$envelope_file"
   # If the envelope file exists, we've already sent it
   if [[ -f $envelope_file_path ]]; then
@@ -37,7 +43,7 @@ send_event() {
   local file_length=$(wc -c <$log_file | awk '{print $1}')
 
   # Add header for initial envelope information
-  $jq -n -c --arg event_id "$event_hash" \
+  $jq -n -c --arg event_id "$fingerprint_value" \
     --arg dsn "$SENTRY_DSN" \
     '$ARGS.named' >"$envelope_file_path"
   # Add header to specify the event type of envelope to be sent
@@ -59,10 +65,18 @@ send_event() {
       --argjson stacktrace "$stacktrace" \
       '$ARGS.named'
   )
+
+  # It'd be a bit cleaner in the Sentry UI if we passed the inputs to
+  # fingerprint_value hash rather than the hash itself (I believe the ultimate
+  # hash ends up simply being a hash of our hash), but we want the hash locally
+  # so that we can avoid resending the same event (design decision to avoid
+  # spam in the system). It was also futzy to figure out how to get the
+  # traceback in there properly. Meh.
   event_body=$(
     $jq -n -c --arg level error \
       --argjson exception "{\"values\":[$exception]}" \
       --argjson breadcrumbs "{\"values\": $breadcrumbs}" \
+      --argjson fingerprint "[\"$fingerprint_value\"]" \
       '$ARGS.named'
   )
   echo "$event_body" >>$envelope_file_path
@@ -203,8 +217,7 @@ cleanup() {
 
     # Only send event when report issues flag is set and if trap signal is not INT (ctrl+c)
     if [[ "$REPORT_SELF_HOSTED_ISSUES" == 1 && "$1" != "INT" ]]; then
-      local event_hash=$(echo -n "$cmd_exit $traceback" | docker run -i --rm busybox md5sum | cut -d' ' -f1)
-      send_event "$event_hash" "$error_msg" "$traceback_json" "$breadcrumbs"
+      send_event "$cmd_exit" "$error_msg" "$traceback" "$traceback_json" "$breadcrumbs"
     fi
 
     if [[ -n "$MINIMIZE_DOWNTIME" ]]; then