`/user/me` endpoint slow response time

[leoraba]

Description

The GET /user/me endpoint is used throughout the data submission workflow to verify each user’s permissions, and it is called in parallel by multiple services during this process, and each response takes a considerable amount of time.

Details

For local testing, a Bash script is run in multiple terminals at the same time to concurrently call the /user/me endpoint.
Executed on a mac connected to the OICR VPN on Tue Apr 21 11:57:10 EDT 2026

SERVER="${SERVER:-https://your-server.example.com}"
BEARER_TOKEN="${BEARER_TOKEN:-your-bearer-token-here}"
SERVICE_TOKEN="${SERVICE_TOKEN:-your-service-token-here}"
SERVICE_ID="${SERVICE_ID:-your-service-id-here}"
REQUEST_COUNT="${1:-${REQUEST_COUNT:-100}}"

TOTAL_TIME=0

for i in $(seq 1 "$REQUEST_COUNT"); do
    echo "Request #$i"
    RESPONSE=$(curl --location "${SERVER}/user/me" \
    --header "X-Service-Token: ${SERVICE_TOKEN}" \
    --header "X-Service-ID: ${SERVICE_ID}" \
    --header "Content-Type: application/json" \
    --header "Authorization: Bearer ${BEARER_TOKEN}" \
    -s -o /dev/null -w "HTTP Status: %{http_code} | Time: %{time_total}s\n")
    echo "$RESPONSE"
    REQUEST_TIME=$(echo "$RESPONSE" | grep -oE '[0-9]+\.[0-9]+')
    TOTAL_TIME=$(echo "$TOTAL_TIME + $REQUEST_TIME" | bc)
done

AVG_TIME=$(echo "scale=6; $TOTAL_TIME / $REQUEST_COUNT" | bc)
echo ""
echo "--- Summary ---"
echo "Total requests: $REQUEST_COUNT"
echo "Total time:     ${TOTAL_TIME}s"
echo "Average time:   ${AVG_TIME}s"

Bellow the result in dev, staging and prod.

Environment	Avg time
Dev	0.69s
Staging	2.08s
Prod	1.52s

Conclusion: Staging and prod takes 1.5 - 2s to respond each request, which for a frequently consumed endpoint it negatively impacts the performance of the submission workflow

[daisieh]

A change has been pushed to staging, if you could give that a check when it is up and running again. Can you see if your tests give you any better results?

The same change is already deployed on dev.

[leoraba]

Tested in staging avg response time is 1.68s, and in dev 0.69s

There is a small improvement in staging, and no change in dev.

[kcranston]

Devops team to investigate difference between the two clusters. Note that the dev response time is acceptable, so aiming to get all clusters to the same response time.

[v-rocheleau]

Infra team will investigate resource metrics for authz to see if this is related to CPU/RAM usage for the service.

If nothing fishy shows up on that side, we'll need to profile the service's code to see what parts of the /user/me are causing the delay.