Hi,
Recently, I've been reading the source code of the latest version (8.7.1) of Elasticsearch and I have a question about the log level settings that I can't figure out.
I noticed that the "failed to sync translog" issue in maybeFSyncTranslogs in Elasticsearch/index/IndexService.java is logged as a warning level:
private void maybeFSyncTranslogs() {
if (indexSettings.getTranslogDurability() == Translog.Durability.ASYNC) {
for (IndexShard shard : this.shards.values()) {
try {
if (shard.isSyncNeeded()) {
shard.sync();
}
} catch (AlreadyClosedException ex) {
// fine - continue;
} catch (IOException e) {
logger.warn("failed to sync translog", e);
}
}
}
}
but in Elasticsearch/index/shard/IndexShard.java, the same issue in createTranslogSyncProcessor is logged as a debug level：
private static AsyncIOProcessor<Translog.Location> createTranslogSyncProcessor(
Logger logger,
ThreadContext threadContext,
Supplier<Engine> engineSupplier
) {
return new AsyncIOProcessor<>(logger, 1024, threadContext) {
@Override
protected void write(List<Tuple<Translog.Location, Consumer<Exception>>> candidates) throws IOException {
try {
engineSupplier.get().ensureTranslogSynced(candidates.stream().map(Tuple::v1));
} catch (AlreadyClosedException ex) {
// that's fine since we already synced everything on engine close - this also is conform with the methods
// documentation
} catch (IOException ex) { // if this fails we are in deep shit - fail the request
logger.debug("failed to sync translog", ex);
throw ex;
}
}
};
}
After reading the context and comments of the latter, I'm wondering if the debug level in the latter should also be changed to a warn level? Or does the use of debug have a different explanation here?
thanks