Why doesn't it just work??

I work for a MSP that manages many customer networks. Despite Solarwinds not being the proper tool for multi-tenancy monitoring, we have done a pretty good job with it so far. There is one thing that bothers me though, every time Soalrwinds comes out with a new feature, it seems like they overlook a few things and it doesn't work the way its supposed to. We have been on an initiative to automate as much stuff as we can to handle our increasing work load, but I have been getting nothing but roadblocks with Solarwinds.

My idea was to take a list of IP address and create a script using the orionSDK to automate loading the devices into the platform. In theory it should work fine, but in reality; it doesn't. Here are the issues that I have run into. If anyone has any viable work-arounds; I'm all ears.

Adding Devices

1. Adding devices through the API using swis.create('Orion.Nodes', **props). This should essentially be the same thing as adding a single node through the web UI. You supply the IP, and SNMP creds and it gets added, but it doesn't work like that. For some reason, Solarwinds decided it would be fun to make us search for, and manually configure all the default pollers that would normally get automatically added when doing it through the UI. Why do I have to supply that through the API when it is automatically detected when adding a node through the UI?? The functionality should be the same regardless of how I'm adding a device. We support a multitude of device vendors, and if I'm going to build a dictionary or database table with all the default pollers for each vendor so my script can add the correct ones, then I may as well just build my own NMS...

2. Adding devices using the discovery API. For this, there are 2 options:

Option A: Setup and run the discovery, but don't import the nodes automatically. This defeats the purpose of automation as we will then have to go into the UI and find the discovery results and finish adding interfaces/volumes. This just saves 8 clicks of the 'Next' button when having to manually configure the discovery.

Option B: Setup and run the discovery, and automatically import the nodes. When reading through the documentation, this seemed like it would be the best way to go. In my mind, it should discover the devices and apply the default pollers, then add them to the system with no interfaces or volumes, we could then add specific interfaces using the AddInterfacesOnNode verb. But, as with everything else I have tried, this doesn't work as expected either. When I run the script to add just the nodes (there is no PluginConfigurationItem configured for the Interfaces in my script), it discovers the device, and brings in all Up/Up interfaces as well as every Volume (including A:). Once again this defeats the purpose of automation because now I have to go back to every node I create and remove the resources I didn't need. The options for configuring the Interfaces PluginConfigurationItem are the same options that you get when configuring the discovery through the web. It seems like it would work until you realize that the filters for interface names are treated as AND instead of OR which basically keeps any interfaces from matching your filters and you end up with nothing.

Creating User Accounts

Here is another feature that doesn't work like it should. I have successfully created a script that can create a list of user accounts. There are 2 problems with the API though:

1. All attempts at creating/applying account limitations to the user accounts through the API have failed, so I still have to manually add them after the accounts are created.

2. While the accounts appear to get created successfully, They are missing the options to set which menu bars are visible to the account. The options don't appear in the account properties until the account has logged into the Web UI at least once. This does not apply to accounts created through the UI, those are fine. In order to make this 'feature' remotely useful (we often get asked to create 10-15 accounts per customer) I had to make another method that logs each user into the web so that I can configure their menu bars. I ended up using Google's Selenium package in python to accomplish this, and while it works; its something that I shouldn't have had to do at all.

There seems to be a big disconnect between UI functionality, and the functionality/efficacy of the API. It makes me wonder if the developers have ever had to use their platform at all, or fully test the features that they implement in the API.

Find more posts tagged with

swis

devices

Python

pollers

accounts

sdk

issues

Accepted answers

All comments

dsimpkins

I feel your pain on this, we are in a similar situation and the goal is automate automate automate!

Have hit the same road blocks as you although found some ways to work around them especially for things like the account limitations.

Happy to share what we have it to help fill in each others gaps if you like?

rschroeder

Your focus is accurate, and I agree with your every point. I'd love for my SolarWinds to be more automated and full-featured. To your list, I'd also include:

Automatic setup of device by category (switch, router, etc.). I'm talking monitoring of them, but their actual config setup, through NCM, leveraging Remediation Scripting and Compliance checking.
Automatic monitoring of all devices
Automatic tie-ins to other modules/products via API using simple radio button selection (or dare I hope for auto-detection of third-party apps). I'm talking about having SW automatically tie into Infoblox, Splunk, etc.
Automatic avoidance of monitoring interfaces/elements that would be selectable via radio buttons, or via imported script, or via simple-or-complex discovery (e.g.: Interface Alias or name does not contain "Controlled")
Automatic Cisco ASA Advanced Monitoring, based on device type discovered. Why should I have to manually check that box when NPM should automatically discover the appliance is an ASA and not a switch or router?
Automatic assignment of new nodes to the closest APE with the fewest elements being monitored.

Yes, all the things I have to do manually today, I'd rather SW did without my input.

Let's put the focus on making our lives simpler, on SW doing the work for us instead of us reacting to every node, volume, or interface as something one-off that needs our individual attention to monitor it, alert on it, or correct it.

lordnykkon

Here is the python script that I have to create the user accounts and log into them so that the menu bar options show up. If you have a way to set the account limitations, I'd love to see it!

#!python3
import re

from selenium import webdriver

from selenium.webdriver.common.keys import Keys

import requests

from orionsdk import SwisClient

import openpyxl

import os

import tkinter as tk

from tkinter import filedialog

import warnings

# import pprint
import time

"""This script works to create the accounts from a spreadsheet, but you need to log into every one of them before it
will give you the option to set the menubars. I created a login function using selenium and Chrome."""

def main():

npm_server = input("Enter Solarinds IP: ")
   username = '********'
   password = '********'

   verify = False
if not verify:

from requests.packages.urllib3.exceptions import InsecureRequestWarning

requests.packages.urllib3.disable_warnings(InsecureRequestWarning)

   swis_npm = SwisClient(npm_server, username, password)

   print("Adding User Accounts:")
   file = get_file()
   accounts = read_file(file)

   for account in accounts:

swis_account = {
'accountID': account['Username'],
'password': account['Password']

   }
   load_npm(swis_npm, swis_account)

   login(npm_server, accounts)

def get_file():

user_dir = os.getenv("USERPROFILE")
   root = tk.Tk() # Creates root window
   root.withdraw() # Hides the root window so that only the file dialog shows up
   file_options = {'initialdir': user_dir, 'filetypes': [('Excel Files', '.xlsx'), ('All Files', '.*')]}
   file_path = filedialog.askopenfilename(**file_options)
   if not file_path:

raise SystemExit # If no file was chosen, exit the program

return file_path

def read_file(file):

warnings.simplefilter("ignore")
wb = openpyxl.load_workbook(file)
sheet = wb.active

lst_accounts = [{
   'Username': sheet['F' + str(row)].value,
   'Password': sheet['G' + str(row)].value,
   } for row in range(2, sheet.max_row + 1)]

   print('\nDone getting accounts from sheet')
   input('\nPush Enter to continue')
   return lst_accounts

def load_npm(swis_npm, props):

uid = props['accountID']

pw = props['password']

   print("Adding user {} to NPM... ".format(props['accountID']), end="")
   results = swis_npm.invoke('Orion.Accounts', 'CreateOrionAccount', uid, pw)
   # print(results)
   time.sleep(2)
   print("DONE!")

def update_account_settings(swis_npm, uid, props):

   print("Updating account settings...", end='')
   result = swis_npm.invoke('Orion.Accounts', 'UpdateAccount', uid, props)
   print(result)
   time.sleep(2)
   print('DONE!')

def login(server, accounts):

driver = webdriver.Chrome()
   base_url = 'http://{}/Orion/login.aspx'.format(server)
   driver.get(base_url)
   time.sleep(1)
   for cred in accounts:

   print('Logging into {}...'.format(cred['Username']), end='')
   url = 'http://{}/Orion/Login.aspx?AccountID={}&Password={}'.format(server, cred['Username'], cred['Password'])
   driver.get(url)
   time.sleep(2)

   print('Done')
   driver.close()

main()