First of all, in ADM100 there is a chapter talking about start/stop SAP instance

For the Win2012 / SQL server environment, instead of running commands like startsap/stopsap, there is an MMC console to perform the task as well

After SAP Starts, the following services shall be running

In the MMC console, the work+disp shall be running

Troubleshooting: If your SAP will not start, check the logs in the work directory first. Sort all the logs by time, and look at the first few of them will normally tell the story

In production systems, you may want to avoid start/stop SAP multiple times, since you may notice in the work directory, there is some .old file. Everytime when SAP start/stop, it will rename all those log files into .old files. However if you stop/start your SAP more than 2 times, the .old files will be overwritten. So the best practice is to back up all the work directory when you try to stop/start SAP. So that you can keep the evidence of previous

As an Example, the system log SM21 is also stored as a file, this file has a maximum size. When the size reach the limit, the retention will start from beginning.

The first interesting thing we found, from troubleshooting point of view, was the SHARK EHP7 server came down suddently. Since we know when the ESXi server was built, the hardware clock was a little odd, which did not sync the time to VMs. So that in the first few days, the time in all Windows 2012 VM was wrong

After we start SAP, from the system log SM21, we found the following

Normally in a production environment, the DB server and application servers are separated. This is for performance, and for high availability as well. In most of the companies there will be a NTP server running in their domain controllers, so do we. So in my past 15 years working as an SAP Basis I have never been able to see this error. So, honestly, this was the first time that I have seen SAP systems down because of "time difference"

To be curious, I ran this ABAP report, after we fix the time issues across our Shark infrastructure. The result shows 'OK'

Another interesting found when I was capturing screenshot this morning: I happen to run SM50 to check the work processes, in my mind the OS files under work directory will always represent these WP trace, in a real time manner. However, this time I guess all these were in SAP memory controlled by SAP kernel, unless someone has actually read the trace, and the file will be written / updated?!

First we check the OS file, take the 3rd work process as an example. At this moment the file size was 0, nothing in there

Then we go to SM50 to check the content of the WP3 trace

Now check the W3 file again, and we did see it's updated with 40kb contents

Hi Adam, which tools you are using to read the log file in your screenshot? I only can open it with notepad which don't have the nice easy read table format. Thanks, Meiping.

Hi Adam, which tools you are using to read the log file in your screenshot? I only can open it with notepad which don't have the nice easy read table format. Thanks, Meiping.

That trace file screen was from SAP transaction SM50 as shown above

System log was transaction SM21.

None of these files are editor friendly though