[TINKERPOP-817] Gryo serialization of large responses fails and causes protocol desync

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.0.1-incubating
• Fix Version/s: 3.1.0-incubating, 3.0.2-incubating
• Component/s: driver
• Assignee: Stephen Mallette
• Reporter: Geoff Reedy
• Created: 28/Aug/15 14:50
• Updated: 11/Sep/15 19:50
• Resolved: 11/Sep/15 19:50

Description

Serialization of large responses in Gryo is not done correctly. It looks like there is some sort of framing or fragmentation issue. I haven't tracked it down the the root cause. It can be reproduced by sending the query "x" * 4066 whereas "x" * 4065 works.

With debug logging turned on, the server indicates that it is writing the correct response message but when it is serialized it comes out only as the single byte 0x01.

The error on the client is, in this case,

io.netty.handler.codec.DecoderException: org.apache.tinkerpop.gremlin.driver.ser.SerializationException: org.apache.tinkerpop.shaded.kryo.KryoException: Buffer too small: capacity: 1, required: 8

but different response sizes give different errors, for example, "x" * 10000 gives

io.netty.handler.codec.DecoderException: org.apache.tinkerpop.gremlin.driver.ser.SerializationException: java.lang.IndexOutOfBoundsException: Index: 118, Size: 0

Issue Links

• is related to: TINKERPOP-814 ConnectionPool can fill with dead Connections