You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
LOGGER.error(KeyValueLogMessage.of("unable to perform no-op operation against fdb",
LogMessageKeys.CLUSTER, getClusterFile()), err);
}
}
However, this is a bit misleading, for a few reasons:
No-ops can fail, and it shouldn't be classified as an ERROR. The most likely reason for this is that when the client is starting up (if using the multi-version client), then a no-op can fail with "cluster version changed".
The error is getting propagated up anyway. The only additional information is maybe what cluster is being used, which may be helpful in a world where different clusters are configured to use different network threads.
We should be able to remove this ERROR-level log and replace it with something that adds the cluster to the error log info. Then clients can decide at what level they want to log this, rather than potentially polluting their logs with (ultimately) spurious errors.
The text was updated successfully, but these errors were encountered:
This removes the `ERROR`-level log on no-op failure. The error was already getting propagated, and that, in practice, should be enough for adopters to detect errors and (if they so desire) log the error. I considered adding logic to include the `cluster` as logging details, but that ended up being a little bit delicate because it only works if the underlying failure is a `LoggableException`, which isn't necessarily the case because we haven't called the error-wrapping logic at that point. So, it was easier to just leave it without that information, and it didn't seem like a huge loss, as the caller should be able to determine that based on the fact that they need an `FDBDatabase` to call this on anyway.
This resolvesFoundationDB#1692.
Right now, we log at
ERROR
if there is a failure when performing a no-op, even though we propagate the error. See:fdb-record-layer/fdb-record-layer-core/src/main/java/com/apple/foundationdb/record/provider/foundationdb/FDBDatabase.java
Lines 482 to 487 in d19f7b2
However, this is a bit misleading, for a few reasons:
We should be able to remove this ERROR-level log and replace it with something that adds the cluster to the error log info. Then clients can decide at what level they want to log this, rather than potentially polluting their logs with (ultimately) spurious errors.
The text was updated successfully, but these errors were encountered: