[
https://issues.apache.org/jira/browse/SSHD-1147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17308533#comment-17308533
]
Alex D commented on SSHD-1147:
------------------------------
Finally I had a possibility to test the SftpClient::openRemoteFileChannel based
approach, using the FileChannel it returns to download the remote file in
chunks, and it worked without any issues against the IBM SFTP.
Though, with a slow link there's a problem with performance when such approach
is used, which totally depends on the buffer size (the bigger it is, the faster
file is downloaded). Downloading the whole file (knowing its size) with one
SFTP operation is always better when it comes to performance.
So, there's no doubt that the patch you suggest will also work.
Maybe it makes sense to introduce a separate "readChunked" method for
SftpClient so that it created {{SftpInputStreamAsync}} without involving the
"stat" operation.
At least it devs will have a choice, as the "read" method itself is much more
handy in use.
> Oh well. IBM's the mountain, it won't move...
True that, and even if they will make a change on their side to allow a
"concurrent" SSH_FXP_STAT operation when the remote file is already open, there
still will be older IBM server versions in use.
> SftpClient is not able to download file from proprietory SFTP servers (IBM)
> with a one time download policy
> -----------------------------------------------------------------------------------------------------------
>
> Key: SSHD-1147
> URL: https://issues.apache.org/jira/browse/SSHD-1147
> Project: MINA SSHD
> Issue Type: Wish
> Affects Versions: 2.5.1
> Reporter: Alex D
> Assignee: Thomas Wolf
> Priority: Major
> Fix For: 2.7.0
>
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> Some propretory SFTP servers support a so-called "one time download" policy
> also known as Extractability Count.
> Current implementation of the input stream based read ofile operation
> org.apache.sshd.client.subsystem.sftp.SftpClient::read(String path) uses
> org.apache.sshd.client.subsystem.sftp.impl.SftpInputStreamAsync
> which performs 2 operations in its constructor:
> 1) Open remote file (SftpClient::open)
> 2) Calculate remote file size using file's handle obtained in 1)
> As a result the following error is produced on attempt to download a file
> from IBMs SFTP:
> SFTP error (SSH_FX_FAILURE): The message [remote file path] is not
> extractable!
> IBM specifically states that it's not their problem at all, and they are not
> going to fix it.
> According to their flow described here:
> https://www.ibm.com/support/pages/node/1121475?lang=en
> "Any attempt from client to issue other SFTP operations like SSH_FXP_STAT
> (reading attributes) makes server to validate extractability_count > 0.
> If found 0, download would be reported as "not extractable".
> Most likely it's possible to use SftpClient::openRemoteFileChannel and work
> with the obtained FileChannel manually without any additional requests to
> stat remote file size.
> Though it would be a cumbersome approach, as reading remote file in chunks
> not knowing its size in this case will be far from optimal.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
