Maybe a RAM error? I remember Alexey or BrightSilence had a similar problem, when a defective RAM returned corrupted pieces.
You could restart the system and do a memory check.
1 Like
Unfortunately not very likely. I see this error on multiple nodes.
I guess this would likely result in failed audits rather than merely canceled ones.
@Alexey:
Any idea why I am seeing this?
I think "downloaded size (0 bytes) does not match received message size (256 bytes)" means that the repairer asked for 256 bytes and 0 bytes were sent.
So, looks like a timeout again or maybe even a partial hang, because the node was able to respond, but was unable to send even 256 bytes after 15 seconds.
What is the difference to context canceled then?
I do not think that there is a difference, both are considered as a timeout.
This one just have more details, that it was unable to retrieve even the header. So, maybe something wrong with the disk?
I see this on several nodes, so I donāt think so.
1 Like
Then it must be related to this setup.
Please, do not assume, that we are ignoring your feedback! Itās pretty useful.
We just do not have a noticeable amount of the support requestsā¦
So, right now it looks like a local problems, not related to the upgradeā¦
Thus - more details are needed (we have a monitoring).
1 Like
2 posts were split to a new topic: Downloaded size (0 bytes) does not match received message size (512 bytes)
3 posts were split to a new topic: Stream closed by peer
OK
thanks @ifraixedes
have a god day 
It seems that the message upload started/download started have been removed from the info log level.
So it is no longer possible to see the time passe between the request and its fulfillment or failure.
How to oget this back without having to switch to debug log level completely?
It will reduce lot of IOPs too but I am more concerned with losing the available space flag. I posted the issue on github.