Skip to content

Common problems🔗

ascli detects common problems and provides hints to solve them.

Error: "Remote host is not who we expected"🔗

Cause: ascp >= 4.x checks fingerprint of the highest server host key, including ECDSA. ascp < 4.0 (3.9.6 and earlier) supports only RSA (and ignores ECDSA presented by the server). aspera.conf supports a single fingerprint.

Workaround on client side: To ignore the certificate (SSH fingerprint) add option on client side (this option can also be added permanently to the configuration file):

--ts=@json:'{"sshfp":null}'

Workaround on server side: Either remove the fingerprint from aspera.conf, or keep only RSA host keys in sshd_config.

References: ES-1944 in the release notes of 4.1, and the HSTS admin manual section "Configuring Transfer Server Authentication With a Host-Key Fingerprint".

Error: "can't find header files for ruby"🔗

Some Ruby gems dependencies require compilation of native parts (C). This also requires Ruby header files. If Ruby was installed as a Linux package, then also install the Ruby development package: ruby-dev or ruby-devel, depending on distribution.

Private key type: ed25519🔗

Keys of type ed25519 and keys in the OpenSSH format are supported by net-ssh using the Ruby openssl gem (>= 3.2.0): the ed25519 gem is not needed. Only OpenSSH private keys protected by a passphrase require the bcrypt_pbkdf gem, which requires compilation of native code:

gem install bcrypt_pbkdf

If the gem is missing, then loading such a key fails with an error mentioning bcrypt_pbkdf.

If the OpenSSL backend does not support Ed25519, then the following message is displayed:

OpenSSH keys only supported if ED25519 is available (NotImplementedError)
net-ssh requires openssl >= 3.2.0 and a crypto backend with Ed25519 enabled for ed25519 support.

To ignore the ed25519 keys of the default identities (~/.ssh/id_ed25519), set env var ASCLI_ENABLE_ED25519 to false.

In addition, if JRuby is used, host keys of type: ecdsa-sha2 and ecdh-sha2 are also deactivated by default. To activate, set env var ASCLI_ENABLE_ECDSHA2 to true.

JRuby: net-ssh: Unsupported algorithm🔗

JRuby may not implement all the algorithms supported by OpenSSH.

Add the following to option ssh_options:

{"host_key":["rsa-sha2-512","rsa-sha2-256"],"kex":["curve25519-sha256","diffie-hellman-group14-sha256"],"encryption": ["aes256-ctr", "aes192-ctr", "aes128-ctr"]}

For example:

--ssh-options=@json:'{"host_key":["rsa-sha2-512","rsa-sha2-256"],"kex":["curve25519-sha256","diffie-hellman-group14-sha256"],"encryption": ["aes256-ctr", "aes192-ctr", "aes128-ctr"]}'

Error: "SSL_read: unexpected eof while reading"🔗

Newer OpenSSL libraries expect a clean SSL close. To deactivate this error, enable option IGNORE_UNEXPECTED_EOF for ssl_options in option http_options.

--http-options=@json:'{"ssl_options":["IGNORE_UNEXPECTED_EOF"]}'

Error: ascp: /lib64/libc.so.6: version GLIBC_2.28 not found🔗

This happens on Linux x86 if you try to install transferd on a Linux version too old to support a newer ascp executable.

Workaround: Install an older version of transferd:

ascli transferd install 1.1.2

See Single file executable.

Error: Cannot rename partial file🔗

This is an error coming from ascp when it is configured to use a partial file name, and at the end of the transfer, the partial file, now complete, does not exist anymore.

This often happens when two transfers start in parallel for the same file:

  • Session 1 starts for file1, it creates file: file1.partial and fills it.
  • Session 2 starts for file1 (same), it creates file: file1.partial and fills it.
  • Session 1 finishes, and renames file1.partial to file1.
  • Session 2 finishes, and tries to rename file1.partial to file1, but it fails as it does not exist anymore...

By default, ascli creates a configuration file ~/.aspera/sdk/aspera.conf like this:

<?xml version='1.0' encoding='UTF-8'?>
<CONF version="2">
<default>
    <file_system>
        <resume_suffix>.aspera-ckpt</resume_suffix>
        <partial_file_suffix>.partial</partial_file_suffix>
    </file_system>
</default>
</CONF>

In container, this is located in /ibm_aspera.

One possibility to avoid that error is to disable partial filename suffix... But that only hides the problem.

For example, when using the container, override that file with a volume and remove the line for extension. Another possibility is to add this option: --transfer=@json:'{"ascp_args":["--partial-file-suffix="]}' : this overrides the value in config file.

Note

If one relies on --lock-port when using containers to avoid parallel transfers in a cron job, this may be the cause, as lock_port does not lock across containers. Use flock instead.