[SPIGOT-3070] Server crash - watch dog timer - skeleton horse ticking Created: 17/Feb/17  Updated: 17/Feb/17  Resolved: 17/Feb/17

Status: Resolved
Project: Spigot
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Bug Priority: Major
Reporter: Virtual Gladiator Assignee: Unassigned
Resolution: Invalid Votes: 0
Labels: crash, skeleton_horse, watchdog
Environment:

    Minecraft Version: 1.10.2 (tested on 2017-02-17 build of 1.11.2 as well)

    Operating System: Linux (amd64) version 2.6.32-573.18.1.el6.x86_64
    Java Version: 1.8.0_25, Oracle Corporation
    Java VM Version: Java HotSpot(TM) 64-Bit Server VM (mixed mode), Oracle Corporation
    Memory: 774215240 bytes (738 MB) / 1610612736 bytes (1536 MB) up to 6442450944 bytes (6144 MB)
    JVM Flags: 5 total; -Xmx6144M -Xms1536M -XX:+CMSClassUnloadingEnabled -XX:+UseG1GC -XX:MaxGCPauseMillis=50
    IntCache: cache: 0, tcache: 0, allocated: 12, tallocated: 94
    CraftBukkit Information:
   Running: CraftBukkit version git-Spigot-de459a2-51263e9 (MC: 1.10.2) (Implementing API version 1.10.2-R0.1-SNAPSHOT) true


Attachments: File r.-7.-6.mca    

 Description   

Tested also with only Essentials and Multiverse plugins, which were needed to warp to the specific location in the specific world where all the skeleton horses (thunder) were spawning.

Same crash every time. It takes a few seconds, but then commands no longer respond and watchdog eventually kicks in to kill the server. Further attempts to join are met with almost instant crashing of the server, forcing a restore from backups. region_fixer finds nothing wrong in the world. No excess entities, no corrupted chunks, nothing. Full crash report and watchdog crash in the pastebin:

http://pastebin.com/xwze3RAd

 



 Comments   
Comment by md_5 [ 17/Feb/17 ]

Mcregionfixer should pick it up and be able to remove entities. Its hard to tell because you've merged two crash logs into one paste

Comment by Virtual Gladiator [ 17/Feb/17 ]

Nibbler! Get out of there!

Are there any updated tools (Linux server-side, like region_fixer.py) for scanning and fixing this type of corruption, or would we have to pull this region into MCEdit and attempt a repair there?

Comment by md_5 [ 17/Feb/17 ]

Chunk corruption ccrash: java.lang.IllegalArgumentException: ChunkNibbleArrays should be 2048 bytes not: 0

And yes, I've seen the second issue before (its a vanilla thing I think) where horses get excess velocity and cause the server to lag and crash.

Comment by Virtual Gladiator [ 17/Feb/17 ]

@md_5

Why do you think the map is corrupted? What tool did you use to come to that determination? region_fixer.py seems to have a different conclusion, which can be seen in the posted comment above. No corruption of any kind detected.

Excess velocity? They are practically standing still. Did you try to load the region file on a server to test it out? They are all just milling around, not doing much of anything.

Comment by md_5 [ 17/Feb/17 ]

Corrupt map and horse entity with excess velocity.

Comment by Virtual Gladiator [ 17/Feb/17 ]

(version: 0.2.1)

############################################################
#################  Scanning world: world  ##################
############################################################

World info:
There are 9 region files, 18 player files and 6 data files in the world directory.

-------------------- Checking level.dat --------------------
'level.dat' is readable

---------------- Scanning UUID player files ----------------
Scanning: 18 / 18 100% ######################################################################################################################################################################## Time: 00:00:00

------------- Scanning old format player files -------------
Info: No files to scan.

---------- Scanning structures and map data files ----------
Scanning:  6 /  6 100% ######################################################################################################################################################################## Time: 00:00:00

------------------ Scanning region files -------------------
Scanning:  9 /  9 100% ######################################################################################################################################################################## Time: 00:00:02

############################################################
################# Scan results for: world ##################
############################################################

Unreadable player files:
No problems found.

Unreadable data files:
No problems found.

Chunk problems:
No problems found.

Region problems:
No problems found.

Comment by Virtual Gladiator [ 17/Feb/17 ]

Crashing region file attached. Just add it to your world, and then teleport to these coordinates or create a warp:

x: -3293.806399738378
y: 78.5
z: -2682.236642669029
yaw: -178.64977
pitch: -1.8000675

Once you tp or warp, the server will stop responding to commands within seconds, then crash a few seconds later on the watchdog timer.

Start walking towards the building in front of you. The horses spawn around the outside. When I test repeated the crash, I walked forward towards the building and around the right curved side. Spigot crashed within 30 seconds.

Generated at Tue Apr 22 04:30:15 UTC 2025 using Jira 10.3.5#10030005-sha1:190c783f2bd6c69cd5accdb70f97e48812a78d14.