Record once on line, SS service appears large area cannot respond to request: Java.io.IOException:Broken pipe

May 26, evening, I talk about the new features deployed online after two days, the interface service can not respond to error characteristics:

1. Log information:

EXEC-147] Com.xpx.middle.web.VproductSearchController-òì³£ðåºjava.io.ioexception:broken pipeòì³£¶ñõ»£º  Org.apache.catalina.connector.OutputBuffer.realWriteBytes (outputbuffer.java:393)  Org.apache.tomcat.util.buf.ByteChunk.flushBuffer (bytechunk.java:426)  Org.apache.catalina.connector.OutputBuffer.doFlush (outputbuffer.java:342)  Org.apache.catalina.connector.OutputBuffer.flush (outputbuffer.java:317)  Org.apache.catalina.connector.CoyoteOutputStream.flush (coyoteoutputstream.java:110) Com.fasterxml.jackson.core.json.UTF8JsonGenerator.flush (utf8jsongenerator.java:1020)

There are no other business error messages.

2. New requests are not fully responsive.

Solution:

1. View CPU memory, the number of TCP connections are not many, MySQL SQL is also normal.

2. See Jstack-l PID also has no very special information.

Finally, the interface prints the node log, calculates several key steps of execution, time consuming, and finds clues.

Summary: Here to say that the interface timeout will cause the entire service is not available, do a good job of interface-dependent services error handling is critical:

FailFast: Quick Find errors

Failover: Failed to switch.

Finally: Change the service read timeout to allow service providers to optimize