Doing show interface takes longer than the the amount of time paramiko waits so it fails expecting a prompt.

Let's find a better way to do this

The shell clients has command and connect timeout attributes that could be controlled by the user

However, the switch factory does not take these params nor it passes them to the adapters factories

That might be useful

enter supports lists but not exit. It'd be great to have that feature :)

I stumbled upon a bug on the cisco 3750.

Sometimes you try to add an ip address to an interfave vlan, it's gonna tell you that it is already assigned to another vlan, you show that vlan's config and it's empty.
To solve : assign the ip to the other vlan, and remove it again and then everything works

It would be AWESOME if netman could handle this bug and make it seems seamless.

Don't know how to reproduce though, here's an example:

SWITCH(config)#int vlan 200
SWITCH(config-if)#ip address 1.1.1.1 255.255.255.248
% 1.1.1.1 is assigned as a secondary address on Vlan300
SWITCH(config-if)#do show run int vlan 300
Building configuration...

Current configuration : 236 bytes
!
interface Vlan300
 description XXXX
 no ip address
 no ip redirects
 no ip unreachables
 no ip proxy-arp
 load-interval 30
end

SWITCH(config-if)#int vlan 300
SWITCH(config-if)#ip address 1.1.1.1 255.255.255.248
SWITCH(config-if)#no ip address 1.1.1.1 255.255.255.248
SWITCH(config-if)#int vlan 200                             
SWITCH(config-if)#ip address 1.1.1.1 255.255.255.248   

... worked!

As tested on a physical switch, when setting an interface on trunk mode, we need to call: switchport trunk encapsulation dot1q

co-dr2.staging(config-if)#switchport mode trunk
Command rejected: An interface whose trunk encapsulation is "Auto" can not be configured to "trunk" mode.
co-dr2.staging(config-if)#switchport trunk encapsulation dot1q
co-dr2.staging(config-if)#switchport mode trunk
co-dr2.staging(config-if)#

On any operation on an interface vlan, brocade enables it.

That was originally to work around them being disabled by default, but it's clearly a problem if your interface vlan is shutdown and you want to modify it with bringing it up

This is related to bug ncclient/ncclient#121

The per-file header license is hard to keep up-to-date and pollutes diff. The license's year of modification should not be a burden. The exact timestamp and author of modification is tracked using git anyway.
I think we should remove those License headers in our open-source project.

Is there anyone against the removal of those headers ?

Interesting talks on the subject :

• Apache 2.0 license says we SHOULD (not MUST) add the license headers.
• Openstack concedes it's not needed.
• Get Rid of Source Code Templates.

A test using a real double proxy should be writen

The proposed way is to launch a netman in the tests and use it as a proxy in a call to the actual flask test

Title says it all! I believe we aren't that far off :)

A juniper with no configuration on a port will not show it when querying interfaces via the netconf protocol

So the /interfaces should look for an official list

There is a CLI command : show interfaces terse
That will show all interfaces names, we need to find how to do it the netconf-way

It seems that when both an IP and IPV6 vrrp groups are defined, netman fails on the get_vlans method with the following error:

ERROR: 'NoneType' object has no attribute 'priority

Getting the vlans on a router that have many vlans will ultimately timeout after 5 minutes. Since netman is only relying on the prompt to determine if a command is done, the router may still be outputting the vlans while netmans pulls the plugs.

Netman could look if there are new characters incoming in the buffer before calling it quits.

When removing a vlan the adapter should clean all its usages that should not exist anymore since the vlan doesn't exist anymore

This is fairly new... here's a sample of the return of the /interfaces on a Juniper

{
"access_vlan": null,
"trunk_vlans": [],
"bond_master": null,
"trunk_native_vlan": null,
"shutdown": false,
"port_mode": "access",
"name": "\nae0\n"
},

The doc is broken on readthedocs for a while.
It's not updated ... this should be fixed

actually the connect to the switch is made before acquiring the lock, this means if 10 sessions opens on a switch with max 5 connections, even though we have a lock, we'll fail.

This mecanism has to be rethought

When a ValueError occurs (such as int('test')), no stack trace is available.

It would be useful to display if the port speed is 1GB or 10GB

Since paramiko 1.15.2 ssh client transport fails with SSHException: Invalid packet blocking. This issue was not present in the previous version. This issue is only present on mac osx

Werkzeug will deprecate the request's data attribute soon in favor of the get_data method. When including netman in a project that has an older version of werkzeug, the get_data method is not found.

We saw Netman wait forever when we display an interface configuration and the uplink port channel is huge.

Command

show running-config interface port-channel 10

CoS

• Netman doesn't wait forever when the switch asks to hit a key to continue the pagination

The following methods have been identified as violating the nomenclature guidelines. Proper compliance tests for those would help us move forward.

$ grep 'TODO' -A1 netman/core/objects/switch_base.py
    # TODO(mmitchell): Should be called unset_access_vlan
    def remove_access_vlan(self, interface_id):
--
    # TODO(mmitchell): Should be called set_interface_state
    def shutdown_interface(self, interface_id):
--
    # TODO(mmitchell): Should be part of set_interface_state
    def openup_interface(self, interface_id):
--
    # TODO(mmitchell): Should be called set_native_vlan
    def configure_native_vlan(self, interface_id, vlan):
--
    # TODO(mmitchell): Should be called unset_native_vlan
    def remove_native_vlan(self, interface_id):
--
    # TODO(mmitchell): Should be called unset_vlan_access_group
    def remove_vlan_access_group(self, vlan_number, direction):
--
    # TODO(mmitchell): Should be called unset_vlan_vrf
    def remove_vlan_vrf(self, vlan_number):
--
    # TODO(mmitchell): Should be called unset_interface_description
    def remove_interface_description(self, interface_id):
--
    # TODO(mmitchell): Should be called set_interface_spanning_tree_state
    def edit_interface_spanning_tree(self, interface_id, edge=None):
--
    # TODO(mmitchell): Should be called unset_bond_description
    def remove_bond_description(self, number):
--
    # TODO(mmitchell): Should be called set_bond_native_vlan
    def configure_bond_native_vlan(self, number, vlan):
--
    # TODO(mmitchell): Should be called unset_bond_native_vlan
    def remove_bond_native_vlan(self, number):
--
    # TODO(mmitchell): Should be called set_bond_interface_spanning_tree_state
    def edit_bond_spanning_tree(self, number, edge=None):
--
    # TODO(mmitchell): Should be called set_interface_lldp_state
    def enable_lldp(self, interface_id, enabled):

Currently, netman depends on ncclient and pulls lxml when installing. This is problematic because lxml often has to be compiled. Additionally, the version of lxml used (3.6.1) clashes with Ubuntu Xenial's packaged version (3.5.0). This library is useless when using netman to access switches other than Juniper models.

A workaround is to specify lxml at the same time as installing netman (or a package requiring netman), as such:
pip install lxml==3.5.0 netman
When doing so, pip honors the specific request for lxml and skips the version requirement for netman.

Using the extra dependencies (https://www.python.org/dev/peps/pep-0508/#extras) or different packages for switch model support (that could all use a "core" netman) would be extremely beneficial for external usage.

Removing the last VLAN of a trunk interface results in an error when commiting. Behavior implemented in internap/fake-switches#37

result :

{
"error": "Vlan None not found"
}

Compliance test should assert error message here:

    def test_fails_when_the_vlan_does_not_exist(self):
        with self.assertRaises(UnknownVlan):
            self.client.get_vlan(2000)

Some brocade use ethe 1/1 in the tagged directive when configuring a vlan and some use only e 1/1

we should support both