Feature: adding hook to execute additional scripts on startup

[williamsbdev]

This will close both #18 and #24.

The scripts are located in the /docker-entrypoint-initdb.d directory.

Paired on this with @benkiefer

[yosifkit]

We will just need to document the limitations of --init-file in the docs

[yosifkit]

LGTM

[williamsbdev]

We added the docs in this PR.

[williamsbdev]

One more thing, we also added the directory creation to the docker file.

[md5]

Doesn't this break if there are no files in $INITDB_DIR with an error like cat: /docker-entrypoint-initdb.d/*.sql: No such file or directory?

[yosifkit]

@md5 is correct, we could just add a if [ -f "$f" ]; then

[williamsbdev]

@md5 @yosifkit @benkiefer Fixed. Thanks for the catch!

[benkiefer]

Just checking to see if this was ready to merge. If there is anything else you need let us know.

[tianon]

The indentation here seems off. Is it using space instead of tabs?

Also, I'm not 100% comfortable with this unless we have some solid documentation for what the exact limitations are. People are inevitably going to run into issues with it, and then come asking why their huge mysqldump file doesn't work properly as-is, since there won't be much in the way of helpful output when it doesn't work.

I think a better stepping stone would be to emulate exactly what Postgres is doing for now, and improve there first. Here's something I've been playing with to implement over there:

for f in /docker-entrypoint-initdb.d/*; do
        case "$f" in
                *.sh)  echo "$0: running $f"; . "$f" ;;
                *.sql) echo "$0: running $f"; gosu postgres postgres --single -E < "$f" ;;
        esac
done

[apolci]

@tianon I just run into one of such limitations (line length limit) with a dump and the effect is that the the DB was initialized and the container was runnging but some tables were missing so it took a while to figure out what was the problem.

[benkiefer]

Can you post your setup so that we can see what you are doing? It sounds like you are doing a dump on container startup?

[apolci]

@benkiefer sorry, I wasn't clear. I just want to confirm the @tianon statement that people will run into issue with this implementation as I did.

My problem was that I was trying to import during the container startup a previously generated dump. Unfortunately the dump contained a pretty long statement (90,000 characters) and the mysql init file is limited to 20,000 characters (http://bugs.mysql.com/bug.php?id=67345 and they won't fix it).
This limit is only for init file run during mysqld startup. The mysql client instead can handle the dump just fine. If the user don't know the difference between running a script with the client and using it as init script for the daemon won't understand easily why the container doesn't initialize properly.

[jgangemi]

is there any movement on this? i could really use the functionality this will provide.

[yosifkit]

I think @tianon was working on a set closer to what we have in the postgres image (to include bash scripts as well as sql), but this will still be very limited and needs documentation to cover the caveats.

[jgangemi]

is there another issue tracking this since you closed this one?

as a work around for anyone else, you could copy the entrypoint script that the container uses and append your init script contents to the end. if you're using mysqldump to create the file, be sure to use --extended-insert=FALSE and --skip-comments

[williamsbdev]

There is not another issue. I was working through the comments but no
progress was made after I addressed the issues they had. I'm leaving it up
to someone else to do this if it is needed and will continue to just use my
fork.

On Saturday, January 24, 2015, Jae Gangemi notifications@github.com wrote:

is there another issue tracking this since you closed this one?

as a work around for anyone else, you could copy the entrypoint script
that the container uses and append your init script contents to the end. if
you're using mysqldump to create the file, be sure to use
--extended-insert=FALSE and --skip-comments

—
Reply to this email directly or view it on GitHub
#31 (comment).

Sent from Gmail Mobile

[jgangemi]

then shouldn't it be enough to provide a link to the mysql docs for the --init-file option in the docs here and also include my comment above about the options to use for mysqldump?

if that's what it takes to get this into the official build, i'd be happy to create a pull request. if there's something else, please let me know and i can look to address it. having this as part of the official image would be extremely helpful to me (and i'm sure others).

[williamsbdev]

Go ahead and try that. LGTM. I gave up as I felt they wanted me to research
how the Postgres was working and change my implementation. I have it
working for myself felt it was sufficient to be merged.

On Saturday, January 24, 2015, Jae Gangemi notifications@github.com wrote:

then shouldn't it be enough to provide a link to the mysql docs for the
--init-file option in the docs here and also include my comment above
about the options to use for mysqldump?

if that's what it takes to get this into the official build, i'd be happy
to create a pull request. if there's something else, please let me know and
i can look to address it. having this as part of the official image would
be extremely helpful to me (and i'm sure others).

—
Reply to this email directly or view it on GitHub
#31 (comment).

Sent from Gmail Mobile