### Bigger Errors - Crashes

Nevertheless, there might be cases where we write some code that **crashes the entire application** making Uvicorn and Python crash. 💥

And still, you would probably not want the application to stay dead because there was an error in one place, you probably want it to **continue running** at least for the *path operations* that are not broken.

### Restart After Crash

 * of {@link java.util.Formatter} specifiers. However, note that if the number of arguments does not
 * match the number of occurrences of {@code "%s"} in the format string, {@code Verify} will still
 * behave as expected, and will still include all argument values in the error message; the message
 * will simply not be formatted exactly as intended.
 *
 * <h3>More information</h3>
 *

      return entry.value;
    }
    entry = cacheEntry2;
    if (entry != null && entry.key == key) {
      // Promote second cache entry to first so the access pattern
      // [K1, K2, K1, K3, K1, K4...] still hits the cache half the time.
      addToCache(entry);
      return entry.value;
    }
    return null;
  }

  @Override
  void clearCache() {
    super.clearCache();
    cacheEntry1 = null;

  // case placement information is lost when resolving the handle.
  //
  // If false, a caller might implement pretty-printing by resolving and
  // iterating over the resulting tensor. This may still be viable if resolving
  // the handle loses information, but `SummarizeValue` would be more precise.
  virtual bool PreferCustomSummarizer() const { return false; }

   * next} field is lazily initialized. The type of that field would be {@code @Nullable T}, and the
   * code would be responsible for populating a "real" {@code T} (which might still be the value
   * {@code null}!) before returning it to callers. Depending on how the code is structured, a
   * nullness analysis might not understand that the field has been populated. To avoid that problem

   *
   * Note that transport-layer success (receiving a HTTP response code, headers and body) does not
   * necessarily indicate application-layer success: `response` may still indicate an unhappy HTTP
   * response code like 404 or 500.
   *
   * @throws IOException if the request could not be executed due to cancellation, a connectivity

 *
 * <p>Valid:
 *
 * <ul>
 *   <li>Live: valid key/value are set
 *   <li>Loading: loading is pending
 * </ul>
 *
 * <p>Invalid:
 *
 * <ul>
 *   <li>Expired: time expired (key/value may still be set)
 *   <li>Collected: key/value was partially collected, but not yet cleaned up
 *   <li>Unset: marked as unset, awaiting cleanup or reuse
 * </ul>
 */
@GwtIncompatible
@ElementTypesAreNonnullByDefault

/// tip

If you need GraphQL, I still would recommend you check out <a href="https://strawberry.rocks/" class="external-link" target="_blank">Strawberry</a>, as it's based on type annotations instead of custom classes and types.

///

## Learn More

    * If you didn't use FastAPI and used Starlette directly (or another tool, like Sanic, Flask, Responder, etc) you would have to implement all the data validation and serialization yourself. So, your final application would still have the same overhead as if it was built using FastAPI. And in many cases, this data validation and serialization is the biggest amount of code written in applications.

Alternatively, we could send the GOAWAY prior to the pod termination (assuming this allows existing and new connections to function still).

## Ztunnel Shutdown/Upgrade/Restart

This could be upgrading to a new version, or changing some part of the DaemonSet spec, etc.

### Ztunnel Shutdown Requirements