Corrupted database backup

[valentinocossar]

Hi guys!

I'm experiencing some trouble with backup:remote function. The exported remote .sql.gz file in /tmp folder is perfect, but when download! command downloads the db dump file, locally I find a 15KB file named db_stage_timestamp.sql.gz that's corrupted.

When I try to extract the file with a double click on it, the Mac's Compression Utility returns me a file with a .sql.gz 2.cpgz extension, and if i try to extract the .sql.gz 2.cpgz file, the Mac's Compression Utility returns me a file with a .sql 2.gz extension and so on.

I also tried to extract it with gunzip command, but returns me that error: gunzip: db_stage_timestamp.sql.gz: not in gzip format

I think that the problem isn't the download! command itself, but maybe the Pathname.new() or something else in the backup:remote function, that's because push and pull (that use the downlaod! command) work well without any kind of problem.

Mac OS X version: 10.11.5

Rbenv version: 1.0.0

Ruby version: 2.3.1p112 (2016-04-26 revision 54768) [x86_64-darwin15]

Here you can find all the gems installed on my Mac:

bigdecimal (1.2.8)
bundler (1.12.5)
capistrano (3.4.1)
capistrano-wpcli (0.1.2)
colorize (0.8.0)
i18n (0.7.0)
io-console (0.4.5)
json (1.8.3)
net-scp (1.2.1)
net-ssh (3.2.0)
psych (2.0.17)
rake (11.2.2)
rdoc (4.2.1)
sshkit (1.7.1)

Any suggestion? Thank you.

[valentinocossar]

Can someone help me? :(

[sacrom]

Hi,

I have found today the same problem (on a linux platform). Following the code I have noticed the problem was on (lib/capistrano/tasks/wpdb.rake) at the "download!" call

   desc "Backup the remote database"
      task :remote do
        on roles(:web) do
          within release_path do
            execute :wp, :db, :export, "- |", :gzip, ">", fetch(:wpcli_remote_db_file)
            download! fetch(:wpcli_remote_db_file), Pathname.new(fetch(:wpcli_local_db_backup_dir)).join(fetch(:wpcli_local_db_backup_filename))
            execute :rm, fetch(:wpcli_remote_db_file)
          end
        end
      end

Which caused the Pathname.write() to eventually being called for every 32k of data or less (overwriting and not appending to previous data). I am not a ruby programmer myself and looking at the Pathname documentation I found no clues. So I changed the previous task to use the File object and it just worked:

      desc "Backup the remote database"
      task :remote do
        on roles(:web) do
          within release_path do
            execute :wp, :db, :export, "- |", :gzip, ">", fetch(:wpcli_remote_db_file)
            download! fetch(:wpcli_remote_db_file), File.join(fetch(:wpcli_local_db_backup_dir), fetch(:wpcli_local_db_backup_filename))
            execute :rm, fetch(:wpcli_remote_db_file)
          end
        end
      end

I don't know if this is the right solution, but just in case it helps somebody...

[valentinocossar]

I've found the (supposed) solution, you've to add .to_s after .join(fetch(:wpcli_local_db_backup_filename)) at the end of this line download! fetch(:wpcli_remote_db_file), Pathname.new(fetch(:wpcli_local_db_backup_dir)).join(fetch(:wpcli_local_db_backup_filename)).

I think this is the right solution, but I'm not sure!

[sacrom]

Yes that's a correct solution as well. The problem I followed led to the net-scp gem where the download! second argument was checked to honour the :write method (see link).

So if the parameter is:

• File -> it just works works,
• String -> A File is created and it works
• Pathname -> Honours the :write but as far as I can see each write overwrites the previous...

So I guess both solutions led to the same result :-)

Thanks for sharing your solution!

[valentinocossar]

Thank you for the explanation and for sharing your solution 😁