How do I increase server memory?
We keep seeing this error:
And when I look at the server settings we have 16GB of physical memory, but only 6GB allocated to the server.
How do I allocate more to the server?
When I click the "Large" button, the memory goes up, but then it won't let me do it because we don't have enough CPUs.
I only want to increase the allocated memory. This is the only process running on this server, so I would like to give it 14GB-15GB.
Thanks!
I have a couple of suggestions that might help:
- For shorter time periods (how short depends on number of purepaths and their average size) try going directly to the Methods Hotspot dashlet and adding all 4 of the Exec columns. You can then sort by the 4 columns (Exec Sum, CPU Sum, Wait Sum, Sync Sum) to see the methods that overall contribute the most.
- For longer periods, use the Layer Breakdown dashlet and maybe even make a copy of it so you can change the chart type, toggle the measure from Exec to CPU, make certain APIs visible/invisible, etc. You can then find times when certain API layers contributed more than normal and can focus on those time periods (by dragging the mouse from a point in front of the time of interest to one after it. You can then drill into Method Hotspots to see which particular methods are taking the most time.
PurePath data is huge and unstructured and takes a lot of horsepower to parse, so it is advisable to work with sort time periods, filters, etc, in order to reduce the amount of PurePath data being parsed. Option 1 deals directly with PurePath data (so keep the time frame short and/or filter in advance of drilling down). Option 2 deals with measures and long time periods are not a real issue.
Also, you can define your own APIs to refine the level of granularity possible. For example, com.example would have the API Example by default. You could make Example Business, Example DataAccess, Example Service or whatever levels of "sub packages" makes sense. These new APIs can then be charted in the layer breakdown or similar charts.
HTH,
dave
Thank you! I will give it a shot.
THANKS!
To limit the Web Requests dashboard to the slowest 20 (or whatever), right-click on the dashlet tab (not dashboard) and select Edit Filter ...
Scroll the tabs to the right and select "Top X". Set the checkbox for "Enable rating-based filtering", set the number of web requests you want to see and how you want them sorted. For 'slowest' select "Exec Max (ms)" although you can also select (e.g.) count to get the most frequent, or CPU Sum to get the biggest consumer of CPU time.
This does not reduce the work required to generate the dashboard, since the Dynatrace Server must still go through all of the web requests in the selected time period to find the "top X".
Several, but not all, dashlets have this same feature.
-- Graeme
Ok, based on this approach, I created a business transaction with all response times > 7 seconds (what I'm looking for) and then filtered by that business transaction. It ran for the 7 days and didn't run out of memory, so this seemed to work. If not, I will report back here. 
When you create a BT, it only works with future data and not with any historical data, so your 7 day filter is probably not giving you what you expect, but in 7 days from now, it should.
Okay, so if I can filter to the slowest 20 or 50 or whatever, that would totally help. So I think that is a great add. How do I set that?
Well, the dashboard to start with to identify the slowest methods is the Response Time Hotspots Dashboard. You can drill down from there.
The dashboard to use for slowest response time is the Web Request dashboard. Note that you can set it to only show you the slowest 20 or 50 web requests.
But that is PurePath data. You can increase server memory by changing the radio button to (e.g.) large, ignoring the warning and hitting the apply button. You'll be prompted to restart the server. This is now the only way to set server memory – you can't set it to a specific value.
Make sure you monitor the server health (e.g., by setting the admin email address so someone gets server alerts) so you'll know if this configuration goes sideways.
-- Graeme
D,
In general you shouldn't be using PurePath data for long-term (e.g., monthly) reporting. You should instead use summary data from the Performance Warehouse, which means either charting or using the Business Transactions Hotspots dashboard.
Can you tell us something about the kind of long-term reporting you are trying to do?
-- Graeme
Yeah, I've read that. 
I just think "Hey, I have more memory, so use it."
Okay, so here's maybe a better question:
How do I find the slowest methods and/or response times for the last 7 days? Basically I want to find the top slow methods, so we can fix them. I want to do it across 7 days because our system use and load varies by day.
Thanks!
D,
I recommend that you run the Deployment Guide to check exactly what resources you need for your Dynatrace Server environment.
Also, what release are you running? Different releases have slightly different memory / CPU tradeoffs.
Finally, you might take a screenshot of the Server Health dashboard (from Start Center --> Monitoring) and post it here.
-- Graeme
Thank you for the quick response!
We are using 6.1.0.8054.
The server health is fine I think:
(The warning is not related to the memory)
The out of memory error happens when we do analysis over long periods of time, and as I understand it from reading the forums, that happens. So people try to limit down the time period, which I don't want to do, especially if I have 16GB available, but Dynatrace is only using 6GB. I just want to give the server more memory, as we have plenty to give,
Thanks again!
