a9e95bc05c
Simplifies and modernizes the go build process. Old vs new build times (without cache): - frontend: 185s vs 33s. - productcatalog: 190s vs 30s. - shippingservice: 200s vs 36s. - checkoutservice: 129s vs 27s. Signed-off-by: Ahmet Alp Balkan <ahmetb@google.com> |
||
|---|---|---|
| .. | ||
| genproto | ||
| .dockerignore | ||
| Dockerfile | ||
| README.md | ||
| genproto.sh | ||
| go.mod | ||
| go.sum | ||
| products.json | ||
| server.go | ||
| server_test.go | ||
README.md
productcatalogservice
Run the following command to restore dependencies to vendor/ directory:
dep ensure --vendor-only
Dynamic catalog reloading / artificial delay
This service has a "dynamic catalog reloading" feature that is purposefully
not well implemented. The goal of this feature is to allow you to modify the
products.json file and have the changes be picked up without having to
restart the service.
However, this feature is bugged: the catalog is actually reloaded on each
request, introducing a noticeable delay in the frontend. This delay will also
show up in profiling tools: the parseCatalog function will take more than 80%
of the CPU time.
You can trigger this feature (and the delay) by sending a USR1 signal and
remove it (if needed) by sending a USR2 signal:
# Trigger bug
kubectl exec \
$(kubectl get pods -l app=productcatalogservice -o jsonpath='{.items[0].metadata.name}') \
-c server -- kill -USR1 1
# Remove bug
kubectl exec \
$(kubectl get pods -l app=productcatalogservice -o jsonpath='{.items[0].metadata.name}') \
-c server -- kill -USR2 1
Latency injection
This service has an EXTRA_LATENCY environment variable. This will inject a sleep for the specified time.Duration on every call to
to the server.
For example, use EXTRA_LATENCY="5.5s" to sleep for 5.5 seconds on every request.