How to get the impala query output log into a vari

问题:

I'm creating unix shell script to execute the impala query. I need to get the output log of impala query. For example I tried the below.

output_log = echo $(impala-shell -i $node -q "select name from impaladb.impalatbl" -o output_file)

Output:

+--------+
| name   |
+--------+
| tom    |
| mike   |
+--------+

Fetched 2 row(s) in 0.83s

Here I'm getting the two name output in both output_file and output_log. But I need the "Fetched 2 row(s) in 0.83s" log in output_log variable. How can I get it?

回答1:

I'm not familiar with impala, so I'm not confident that what you are doing is the most efficient way to query impala. However, you're trying to focus in on a specific line of output; that I can answer.

There are many ways to do this. Maybe the most straightforward is grep:

output_log = echo `impala-shell -i $node -q "select name from impaladb.impalatbl" -o output_file | grep Fetch`

回答2:

Try this:

Solution 1:

output_log=$(nohup impala-shell -k --ssl -i $node --verbose --delimited --query="select count(*) as cnt from impaladb.impalatbl" 2>/dev/null)
echo $output_log

Solution 2:

output_log=$(echo `impala-shell -k --ssl -i $node --verbose --delimited --query="select count(*) as cnt from impaladb.impalatbl" -o output_file | head output_file`)
echo $output_log

回答3:

I solved the problem. The way it works is that impala sends the query output in a different stream and the other information w.r.t the query in a different stream.

hence all you have to do is

impala-shell -i $node -q "select name from impaladb.impalatbl" 2>output_file

the 2> will send the output containing the "Fetched 1 row in 10seconds" in the output file. Now you can grep it or do whatever you want .

Suppose you want to store the same output in output_log variable , then use

 output_log=$(impala-shell -i $node -q "select name from impaladb.impalatbl" 2>&1)

here 2>&1 will send the output to stdout which will be used for assinging the value to variable.

For more information on this , just put 2>&1 in google search and learn more about it.

Hope it helps you!!

Some additional observations

2>&1 redirect output from stderr to stdout but stdout also gets the query output, so when you store it in a variable , it will get query output as well as the extra information like"fetched 1 row in 3seconds"

but

when we use

2>a.txt then only the stderr output is getting redirected. so a.txt will contain only information like "starting impala.....fetched 1 row in 2 seconds" . and then you can grep from file and put that in a variable.

just wanted to highlight this slight difference I observed between storing in a file and storing in a variable.