I want to zip the sql dump if and only if mysqldump does not produce any error. I try to put the wrong password, but it still perform the gzip.
mysqldump -u username -ppassword dbname |& if [ $? == 0 ]; then gzip > test.gz; else echo "error"; fi What is wrong with my command or is there any better solution?
I think you could remove |& (the reason being you do not need to pipe either the stdout or stderr into the next conditional, if anything you can just use ';' as a command separator). Something like this should probably work:
#!/bin/bash mysqldump -u myuser -p mypasswd > mydb.dump if [[ $? -eq 0 ]]; then gzip mydb.dump else echo >&2 "DB backup failed" exit 1 fi Edit: to check the success of the gzip you could do something like this:
mysqldump -u myuser -p mypasswd | gzip > mydb.dump.gz && echo "success" || echo "failure" However this would report success even if mysqldump failed, i.e. if the backup failed for most reasons other than gzip filling up the disk.
You can also
mysqldump -someparams dump.sql && gzip dump.sql || echo "Backup failed" Or
mysqldump -someparams dump.sql [[ $? == 0 ]] && gzip dump.sql || echo "Backup failed " A pipe executes the two sides in parallel. The way you're attempting to do this is conceptually impossible: you can't test the status of the mysqldump command until it's finished if you do this test in parallel with the execution of mysqldump. You need to run mysqldump, wait for it to finish, and then decide whether to run gzip.
Since mysqldump has to finish running, its output has to go somewhere. Presumably you expect the output to be large since you're compressing it. Therefore the sensible option is to compress it. So compress the output unconditionally.
mysqldump -u username -ppassword dbname | gzip > test.gz Note that I used |, not |&. Using |& here makes no sense: if there are any error messages, they'd end up mixed with the dump and it would be impossible to restore the dump.
The problem that remains to be solved is detecting whether mysqldump succeeded. Assuming that this is a bash or ksh script (i.e. it begins with #!/bin/bash or #!/bin/ksh or the like, not with #!/bin/sh), set the pipefail option, so that the pipeline fails if any part fails. (By default, the status of a pipeline is the status of its rightmost command and the other commands' status is ignored.)
#!/bin/bash set -o pipefail -o errexit tmp="mydump.tmp.$$.gz" trap 'rm -f "$tmp"' ERR INT TERM HUP mysqldump … | gzip >"$tmp" mv "$tmp" mydump.gz Setting the errexit option ensures that if the pipeline fails then the script exits at that point (with the same error status as the pipeline). Thus a file called mydump.gz is only created if the dump was successful. The trap command sets up a trap so that if the script fails or is killed by one of the listed signals then the temporary file is deleted.
date && set -o pipefail -o errexit && mysqldump --compress --single-transaction db_name | gzip > "db_name.sql.gz" && date. Without setting the bash options, the date call at the end of the chain will run; with the bash settings, the date call will not run, which is what OP wants. You could also use bash build-in table $PIPESTATUS[x] to trap errors.
for example :
$ cmd1 | cmd2 | cmd3 | .. and so on..
Then you'd see which command has failed
If you want to avoid use of a temporary file another solution is to delete the resultant file on error:
( mysqldump ... || rm -f test.gz ) | gzip > test.gz Or, if you want to have an explicit flag, something like this:
rm -f test.ok ( mysqldump ... && touch test.ok ) | gzip > test.gz if [[ $? -eq 0 && -r test.ok ]]; then echo it worked else echo something went wrong fi 