Kicked from server, Console spams "Netty Epoll Server IO #6/ERROR"

Some observations: let's say player A is in the affected area.
When player B moves into view distance of player A (e.g. using a Nether portal), both are kicked.
When player B logs in within view distance of player A, usually only player A is kicked, but sometimes both are kicked.

This plus the Netty-related exceptions point towards the error being related to the Spawn Player packet.

Well I can confirm a number of things right now: It's not a command block, (if it is, turning them off in server.properties doesn't make any difference) it's not an entity, as erasing them all did nothing, and it's not some ethereal world or chunk corruption, as completely refreshing the world by turning it all into an MCEdit schematic and reimporting it doesn't fix it either.

Unfortunately, that still doesn't tell us what it actually IS. My best guess at this time is either a command that's too long, (I've read they shortened the max command length, and I guess they just keep forgetting to account for now invalid block or tile entity data. >->) or a sign somewhere with either commands or text formatting. (These are almost completely broken in the 1.9 snapshots, and have become rather infamous for bugs similar to this.)

Tomorrow I'll see if I have time, and systematically delete objects we think might be the cause until it stops kicking. The most annoying part about troubleshooting this is the necessity of a second player to do anything at all. >-<

There are no command blocks in our affected world.

For reference, the affected world can be downloaded at http://api.wurstmineberg.de/v2/world/wurstmineberg/backup/latest.tar.gz. The two farms at spawn causing the bug are turned off and can be turned on by flicking the levers at 44 114 -10 and -69 70 -5.

Good, that's significantly less work for me then. What about signs? Or any structures made or placed with clone/fill/setblock? I've read that could potentially be the cause on a similar bug report.

To my knowledge, the world was only ever loaded in vanilla survival. Since we're a snapshot server, there may be associated weirdness, though.

And yes, there are quite a few signs in the area.

The affected area contains signs that have previously been affected by MC-82968 and MC-85627, but another area which also has these signs appears to be unaffected.

Interesting, but probably unrelated. Our server had both of those, but most notable was the first one, which, (as I posted on that report as well) was, at least in my case, actually a problem with either one or more item frames with a custom map, or a small collection of minecarts underground. Deleting all entities within the effected chunks fixed that, but that doesn't seem to help this time.

Not sure if it's at all related, but on today's snapshot, the following error is spammed constantly in the server log in addition to what's been shown in this report:

[16:56:18] [Server thread/WARN]: Tried to assign a mutable BlockPos to tick data...
java.lang.Error: class cj
at ll.a(SourceFile:479) [custom.jar:?]
at ll.a(SourceFile:472) [custom.jar:?]
at ajk.f(SourceFile:117) [custom.jar:?]
at ajk.a(SourceFile:64) [custom.jar:?]
at rb.T(SourceFile:778) [custom.jar:?]
at rb.d(SourceFile:721) [custom.jar:?]
at xj.a(SourceFile:526) [custom.jar:?]
at xj.m(SourceFile:329) [custom.jar:?]
at agq.a(SourceFile:1436) [custom.jar:?]
at ll.a(SourceFile:661) [custom.jar:?]
at agq.g(SourceFile:1414) [custom.jar:?]
at agq.k(SourceFile:1305) [custom.jar:?]
at ll.k(SourceFile:540) [custom.jar:?]
at net.minecraft.server.MinecraftServer.D(SourceFile:625) [custom.jar:?]
at kv.D(SourceFile:314) [custom.jar:?]
at net.minecraft.server.MinecraftServer.C(SourceFile:553) [custom.jar:?]
at net.minecraft.server.MinecraftServer.run(SourceFile:457) [custom.jar:?]
at java.lang.Thread.run(Thread.java:745) [?:1.8.0_45]

Affects 15w36c.

"Tried to assign a mutable BlockPos to tick data" is a different bug - MC-87916.
This bug has been known to kick ALL players on my server. Affects 15w36c.

I've been having the exact same issue on my server with the same output message in versions 15w35e-15w36d. It can only handle one player in an area and boots the other. This happens most often in the spawn chunks, but has happened elsewhere in the Overworld as well. So far the nether seems unaffected. I think it only happens when the spawn chunks are loaded, so perhaps they have some sort of issue? It's really got me worried that I might have to abandon my spawn area.

If it is spawn specifically, I don't know what could possibly fix it then, as deleting the entire region file containing spawn and it's surrounding chunks didn't do a damn thing.

The server always keeps on running just fine, but sometimes it crashes the client of the person disconnected. It's odd.

I think that's just an unrelated Minecraft, Java, or user PC issue. Very often, my client will freeze or crash when disconnecting from a server, even on my own terms. It is worth noting however, that when it happens to another player, it produces an error of it's own in my client side game output of the launcher. Unfortunately, I don't have anyone to test with right now to give you all that report, so if somebody else could give it a go, that would be fantastic.

Alright, at least on my server, I've narrowed it down to command blocks. Not specifically an active command, but rather a command block somewhere that has invalid NBT data. If this is effecting people with survival servers as well, my best guess is that it can just effect any block able to contain text.

I currently have four command blocks on my server, all simple TP commands. They were placed in 1.8, before the recent command block update. I tried both replacing them and removing them, but I still seem to have the same issue.

Did you try removing them in game, or in MCEdit? I did the latter, and if our servers share the same trigger, that seemed to do it for me.

Thanks for the suggestion. I went and tried deleting the whole area with the command blocks in MCEdit. Unfortunately the problem persists for me.

Hmm.. what else is in the area? As much as I hate to say it, I don't think this bug is just going to fix itself, and it's not looking like it's going to get a Mojang assignee anytime soon. We should all probably pool our info to try and at least find a repeatable cause.

It's an area in the middle of the ocean with an iron farm, witch farm, and slime farm, each dispensing their drops into water streams funneling up into one item elevator that goes into an item sorter. It's an isolated area and those are literally the only builds around. So at any given time there's maybe 20-30 items traveling through the system. I have much more redstone circuitry and higher yield farms elsewhere on the server, but they've never been an issue. I can definitely afford to abandon the area, it's just an inconvenience haha.

Are there any signs in the build? I'm just trying to think of any possible constant besides the maps being 1.8 and below in origin. I've been building on my map since beta 1.5, and as a result, there's so much different crap around spawn, that I couldn't even begin to pinpoint it beyond a certain point.

I removed all signs after reading a comment you posted earlier. It didn't seem to change anything for me.

Hmm.. Do you think you could isolate the area and upload it as its own map? I only ask cause I assume it's fairly large, and my internet is kinda slow. I might be able to make some comparisons and try running it through MCEdit or something.

I just did some testing with a friend. If I turned all the farms off, she could stay on. If I turned the witch farm on, which involves tripwire, redstone and piston powered block shifting (about 128 slabs shift back and forth repeatedly), she got booted. If I turned on or off the slime farm, which involves a lot of flowing water updates, she got booted. It seems that whenever the server is getting taxed with block updates it booted her. We tax the server much, much more with high mob and entity counts at other locations(getting the server can't keep up messages and all). But this seems to be more about block updates.

Edit: All these things never triggered this issue for me in 1.8.

That actually narrows it down quite a bit. I'll take a look around my own world tomorrow. Though, how it connects to command blocks in my case is beyond me.

I didn't really think about it before, but I have noticed that I'm getting a lot more "can't keep up" warnings than usual lately. At the time, I just figured it was always doing that, but because I was running Spigot on 1.8, and for some reason, that didn't output to the console. This only started on the first 1.9 snapshot before any of the real game breaking bugs started. For me, all of the real problems started with MC-82968 in 15w33c, then evolved into this after I fixed that.

Well, my server has no command blocks, and kicking has been known to occur far from any farms with redstone. Maybe tile entity updates in general? A witch farm, for example, has hoppers etc. to take loot.

Bloody hell, I figured it out. Idk how useful this info will be to anybody else, but in my case, the culprit seems to have been a redstone/command block contraption of my friend's. What it does is constantly change the block in the walls and ceiling of a small room to a different colour of wool on a minecart clock. Tomorrow, I'll import the contraption into a flatgrass map and see if I can replicate it. If I can, I'll upload the effected map for whenever Mojang assigns somebody to this issue.

Affects 15w37a.

BTW regarding the "environment" in this bug report: This occurs on OS X clients as well (with a server running Debian.)

In our case, the cause is a mob farm at spawn, whose water flushing system causes a heavy amount of block updates every few seconds. After disabling the clock, multiple players can be online at spawn at the same time again.

Just a brief update, I can confirm now that it is definitely something to do with rapid block updates on a large scale. Setting off a bunch of TNT, or using a command block to fill, setblock, or replace more than 50 blocks at a time over the course of a few seconds will cause the error and subsequent kick. It began happening to us again, for instance, when we were blasting out a large chamber underground, and then subsequently converting the walls from stone to packed ice.

Hopefully now that this can be identified and replicated, Mojang will take notice and be able to fix it.

Oh, and for the record, thanks all for the help in finally tracking this sucker down. I've dealt with a lot of hiccups and roadblocks in my days running a server. (It'll happen when you use the same map since beta 1.5! 😛) But this has by far been the most difficult to pinpoint and deal with. You have my sincerest gratitude, all of you. 🙂

Yeah, I would agree with that conclusion. In my case it was the water flow updates and large scale piston shifting. Hopefully they will fix it sooner rather than later.

I still can't pinpoint cause for my crashes with the block updates thing alone! However, it does explain why it crashed in an end city (someone placed a water bucke and the water went all the way down).

A similar issue is occurring on my 15w37a server. Pretty annoying issue... the server seems stable if one person is online but as soon as another player joins - the server goes haywire and lists the following error in the console:

[Netty Server IO #6/ERROR]: io.netty.util.IllegalReferenceCountException: refCnt: 0

The second person that joins also cannot be seen by the player who was already in the server. The second person is completely invisible to the first player but can still interact with the world. Sometimes the server will stay up for up to a minute after the error and then kick any players on the server out, afterwards closing itself.

Also, If you are playing on the server and a second person joins the server and spawns far away from you, teleporting to the second person's location will crash the server instantly.

Hopefully this gets resolved!

Try getting people on your server to disable large farms/avoid placing water buckets to try to reduce the number of block updates.

The invisibility issue is something completely different. I've noticed that as well on my server since fixing this bug. It can usually be temporarily resolved by restarting the server.

I'm running a snapshot-testing server, and having this issue on 15w38a. On a new map where players are generating chunks, we get the netty error spam on the console, and the disconnects from clients.

On a Skyblock-type map where no chunks are being generated, we can have farms going and command block contraptions running with no problems or netty errors on the console.

Combined with the different proximate causes above, it seems like absolutely anything that stresses the server can cause this issue?

Still present in 15w38b. This time it was caused by bonemealing a bunch of oak saplings, creating several large trees very quickly.

Again, more fuel for the block updates camp.

On a pre-generated Skyblock void world (so there's no "hidden" block updates happening, and no redstone or fluid updates happening), growing a single branching oak or 2x2 tree type is enough to cause Netty error spam and disconnects, almost every time one grows.

Yeah, trees cause block updates, and a lot of them. Many block updates in one tick=this bug.

Still present in 15w39b, and 15w39c.

Can also confirm for 15w39c as well as the observed disconnects happening on major block update counts (such as trees) and newly created chunks

Still present in 15w40b.

Affects 15w41b, please reopen.

We can still reproduce this on 15w41b, it is not fixed.

Still an issue on 15w41b, particularly during chunk generation. The server seems to be doing a lot more "running behind, can't catch up" in this snapshot as well.

For the record, MC-89127 is a duplicate of this and has been reopened.