Occasional Issue

Posted : admin On 4/7/2022
-->

Adjective periodic, broken, occasional, recurring, irregular, punctuated, sporadic, recurrent, stop-go (informal), fitful, spasmodic, discontinuous After three hours of intermittent rain, the game was abandoned. Intermittent explosive disorder involves repeated, sudden episodes of impulsive, aggressive, violent behavior or angry verbal outbursts in which you react grossly out of proportion to the situation. Road rage, domestic abuse, throwing or breaking objects, or other temper tantrums may be signs of intermittent explosive disorder. Adjective periodic, broken, occasional, recurring, irregular, punctuated, sporadic, recurrent, stop-go (informal), fitful, spasmodic, discontinuous After three hours.

Most issues discovered with the Microsoft Teams client can be traced back to firewall or proxy connectivity. Verifying that the necessary URLs, IP addresses and ports are opened in your firewall or proxy will minimize unnecessary troubleshooting. For specific information on URLs and IPs required for Microsoft Teams, please see the Microsoft 365 and Office 365 URLs and IP Address support article. The following scenarios require specific URLs and ports to be opened in the firewall.

  • Authentication

  • Microsoft Teams Client Connectivity

  • Collaboration

  • Media

  • Shared Services

  • Third Party Integration

  • Skype for Business Interoperability

  • Skype for Business Client Interoperability

When Teams is offline or in low bandwidth conditions

The good news is that Teams keeps running even when you're offline or running in low bandwidth conditions. Teams saves all your unsent messages for existing chats (for up to 24 hours) and sends them as soon as you're back online. If you're offline for longer than 24 hours, Teams lets you choose to resend or delete unsent messages. We're working on adding this functionality to new chats and will update this documentation when that's available.

Related topics

Diagnosing intermittent “network” problems

There’s that one thing that customers usually ask, and that question is if I would be able to help diagnosing a problem on the network. My answer has two parts:

  1. If we can capture the problem situation in packets, I will find it
  2. When I find it, I’ll tell you if it’s a network problem (which, in my experience of over 10 years, is the case in only about 20%)

The trouble is: there are some problems where it’s not easy to capture packets, and that’s when you don’t know the correct capture location (usually meaning “there’s too many possible locations”) or the exact time for the packets with the symptoms to be recorded – or both.

The easy stuff

When a customer tells me “well, whenever I do this, the problems shows up”, I know that it won’t take long to diagnose the network. If the problem can be reproduced you can easily capture packets, simply because you know where to capture (at the persons PC/Mac/whatever it is) and when (while the person is doing whatever causes the problem). Those “easy” problems are the reason why most experienced analysts I know always carry a portable switch with a static or configurable SPAN port feature or a TAP in their laptop bags at all times. BTW, “hubbing out” by using a network Hub instead is an ancient technique and should be avoided at all costs (except in very special situations where the medium is still half duplex).

IssueIntermittent network problems

A word of warning to those still carrying Ethernet hubs: if I find out about you doing that in switched production networks, I’ll most likely make fun of you in my talks at the next Sharkfest! 🙂

The complicated stuff

If you have an intermittent problem the first thing to decide is where and how to capture packets. This decision is mostly based on what kind and number of capture devices you have, and what the network topology requires. In most cases the capture setup is a balancing act between choosing enough spots to have a high probability of capturing relevant packets and the number of capture devices you can deploy.

Some capture situations allow you to use SPAN ports, but in other cases you have to use TAPs, for example:

  • if proof is required that all packets were in fact on the wire (can’t do that with SPAN because it may omit packets, leading to false results)
  • if timings need to be measured as exact as possible (also requires a professional TAP with a built in high resolution timer, or simple TAP in combination with a professional capture device)
  • if the traffic is too much for a switch to SPAN you may need to deploy filtering TAPs

The most complicated capture setup I ever designed in an enterprise level network analysis situation took about two weeks just to decide which devices to use in which location (I’ve mentioned this one in the first of the VMware capture posts already). It involved a huge virtual server cluster environment running DRS on fully automatic, with hundreds of ESXi servers. Plus there were quite a few network analysis nightmares like reverse proxies, forward proxies, load balancers and NAT gateways. A lot of the communication was SSL encrypted, and of course we had no access to the private keys. It doesn’t get more complicated than that (and we still helped the customer solve the problem).

One thing is pretty much always the same in situations like this: you’ll end up digging through tons of capture files with millions of packets. The biggest job my team ever analyzed had close to one TByte of capture files, in thousands of files. Now, how to analyze something like that? One thing is certain: you can’t look at all those packets in Wireshark, unless you visit the groundhog each day in the morning. You need to drastically reduce the amount of packets you’re going to examine first.

Reducing the number of packets to look at

The best way to minimize the amount of packets to look at you should have as many additional information about the problem occurring as you can, including

  • the IP addresses of the computers/network devices involved in the problem
  • the (more or less) exact time of when the problem showed up, which is not always easy to make really clear to the relevant users that are asked to report it. It’s often a problem to have them realize how many packets can “happen” in a minute, so make sure they report as close to the time the problem occurred as possible (and not like 15 minutes later when they find the time)
  • “beacons” in the capture that tell you where and what to look for, e.g. a ping packet that a user triggered using a custom script when she/he noticed that the problem occurred (usually by clicking on an Icon on the desktop that executes the script)
  • server logs with correct time stamps. That requires both the servers and the capture devices to have accurate system times set, of course.

Now, if you have access to commercial tools that are able to handle large file sets you can use those the help you digging through your captures, but that’s not always an option. Instead the relevant packets are often carved from the huge pile by leveraging tshark and mergecap scripts with read filters, e.g with a batch file containing something like

Issue

First, tshark is used to read all files, then apply a display filter (working like a “read filter” in this case), and write the result to a new file. Afterwards, all the new files are merged together into one more or less big file (in Windows batch mode you need %%a, while on the command line you’d use just %a). The second step may be skipped or modified if the file gets to big to be opened in Wireshark, but it often helps to have as much details in a single file as you can. Keep in mind that the 64 bit version of Wireshark can open larger files than the 32 bit version.

Intermittent Connection Issue

Leveraging TraceWrangler

A new way of dealing with a large set of capture files is to use TraceWrangler. Right now most people knowing this utility think of it as a anonymization/sanitization tool, which is correct – but that’s not all it is supposed to do. Sanitization is only one of the things it can do, but the main focus of TraceWrangler is to help working with multiple files. So if you have a big pile of capture files you could add them to TraceWrangler (you may need the 64 bit version and a few GByte of RAM if you have a lot of packets) and have it scan them:

Issue

Now, open the “Communication Details” window via the “Tools” menu and you’ll get a endpoint and conversation statistic much like the one in Wireshark – but calculated over all files currently added to the file list, like in the following example (1.6 million TCP conversations, calculated over a total of 152 files of about 38GBytes with 59 million frames):

There are two additional columns in the TCP statistic window that you do not (yet?) see in Wireshark:

  1. Initial RTT: this column will show the initial round trip time (if available – meaning if the packets for the three way handshake are available). It’s often useful to sort the conversations by this column to find those that have a rather high iRTT – sometimes more than a minute. Wireshark doescalculate iRTT (since 1.12), but it’s not a sortable column in the conversation statistics yet.
  2. Status: in the status column TraceWrangler will tell you the general status of the conversation. It will not only show if there is a partial of complete handshake and tear down, but also if there are any other topics of interest. E.g. if a conversation spans multiple files, or if the frames have out-of-order timestamps or if you only have one side of the communication. There is a status column in the other tabs as well, but the one for TCP has the most details.

When you find something interesting, you can always use the popup menu to either copy a Wireshark display filter string to the clipboard and apply it in Wireshark, or you can extract one or multiple conversations to a new file (e.g. if you want to know why a conversation might have an initial RTT of 90 seconds):

Occasional Anger Issues

I often leave TraceWrangler open and copy filters over to Wireshark, which is more convenient than picking filters from statistics windows of the same Wireshark instance.

Let’s assume we want to know why a conversation would need 90 seconds for the initial handshake. The easiest way to do that is to mark the TCP conversation and hit “Extract to -> single file”. After extracting the conversation to a new file it’s time to look at the packets in Wireshark:

Intermittent Issue Meaning

Those 12 packets were easily extracted from a huge pile (close to 60 million) with a few mouse clicks. Of course it doesn’t always work as simple as that, but if you have some additional details like I mentioned in the bullet list above you can extract a lot without having to read tons of packets. It’s a good example of a very problematic session, where the client obviously suffers from heavy packet loss (the capture was taken right in front of the server).

Samsung Tv Sound Issue Intermittent

Now for a small challenge: can you tell why TraceWrangler diagnosed a iRTT of 90 seconds from the screenshot above? And is that number correct, or do you think iRTT should be a different value (Wireshark 1.12.2 says iRTT is 6.122721 seconds)? Write a comment 😉