what may cause TTransportException: Transport not open? about happybase

[wbolster]

This happens when you try to use a closed connection. Did you disable autoconnect perhaps? If so, try connection.open() before use.

[saintthor]

thank you. the real mistake is one table is not created in the second hbase. when writing data via batch.send() of the missing table, the error occurs.

i think it would be better to raise a TableNotExistError than a TTransportException here.

[wbolster]

Unfortunately the Thrift API exposed by HBase is not very robust when it comes to error handling and error reporting, so I'm afraid there isn't much HappyBase can do to make this any better on the Python side. :-(

[saintthor]

a new problem, code as:

try:
batch.send()
except Exception, e:
print e
conn.close()
conn.open() # or conn = happybase.Connection( ... )

after running some days, the exception 'Transport not open' keep occurs. when i terminate it and run it again, it turns well.

what may cause this?

i think if hbase get some wrong that cause error and then become fine, conn.close and conn.open should make it work. if they didn't, to restart the process wouldn't either.

[wbolster]

The connection.open() calls into the Thrift library to open a new connection, so I'm not sure whether HappyBase is to blame here. A network glitch may cause failures. Is the stack trace from within the batch.send() or from conn.open()?

[saintthor]

the stack trace is from within the batch.send().

while the trace keep occurs, i connect the same hbase via python shell with no error. then restarted it, all ok.

[saintthor]

the exception also may be IOError(_message='Connection refused').

[wbolster]

I'm quite sure the errors you see are caused by network failures. Does it happen when you run the HappyBase process on the same host as the Thrift server (localhost connection)?

— Wouter

(Sent from my phone. Please ignore the typos.)

saintthor [email protected] schreef:

the exception also may be IOError(_message='Connection refused').

---

Reply to this email directly or view it on GitHub:
#45 (comment)

[saintthor]

we run hbase on about 10 servers. and run 1 thrift on each node.

[wbolster]

Is the Python client running on the same nodes? I suspect this is a network issue between the host that runs the Python process and the host that runs the Thrift daemon.

[saintthor]

no. the python client is on another server.

even it is a network issue, i think it should be fixed by close and open when a new process runs ok.

[wbolster]

Is this reproducible if you also call .close() first? If so, it might be something within the Python Thrift library (outside HappyBase).

[saintthor]

i think i catched the bug. i didn't recreate the batch after reconnecting.
i will test it when the hbase become well. now, it is down.

2013/12/5 Wouter Bolsterlee [email protected]

Is this reproducible if you also call .close() first? If so, it might be
something within the Python Thrift library (outside HappyBase).

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/45#issuecomment-29833444
.

[rrjia]

make sure you have thrift server started before you try to connect.

hbase thrift start -threadpool
Also, hbase/bin/hbase-daemon.sh start thrift

make sure you have thrift server started before you try to connect.

hbase thrift start -threadpool
Also, hbase/bin/hbase-daemon.sh start thrift

Hi, @William-jia When i start thrift server using this command hbase thrift start -c -nonblocking, and i checked jps the thrift server is initiated, why we still need to use hbase-daemon.sh start thrift again?