Just because you can’t see or hear your Amazon EC2 instances the way you can in your local data center, doesn’t mean you shouldn’t be concerned about them. Think of SystemTap as a fabulous window into your VM.
Have you ever wondered what’s going on deep inside your Linux instance? Why some processes run so slowly? Why your disk is thrashing around so wildly? Have you tried to write your own system’s tools and wish that you were able to export or access certain systems’ internal variables or information? Have you tried to write your own Linux kernel module but felt that the learning curve was way too steep?
Let me introduce you to SystemTap.
(The beautiful SystemTap logo was designed by Andy Fitzsimon – who sat next to me when we were both working at Red Hat in Singapore.)
SystemTap is an open source software package to help you analyze or troubleshoot a running Linux instance. It provides a dynamic instrumentation framework that lets you probe a system to gather information and a scripting language translator that interprets scripts written in a C or Perl-like language into C code, that could be compiled into Linux kernel modules. These modules will then be loaded into the system to generate outcome specified by your script.
But enough introduction. Let’s get started already!
Installing SystemTap on your AWS EC2 instance
This article assumes that you know how to create and launch an EC2 instance. If not, I recommend taking Cloud Academy’s Compute Fundamentals for AWS Course as part of the AWS Fundamentals Learning Path.
Connect to the instance using the OpenSSH SSH client.
$ ssh -i "cloudacademy-keypair.pem" ec2-user@[redacted]
Note the Linux kernel version we are using. In this case, it’s version 3.10.0, and the release number is 229.el7.
$ uname -a Linux ip-[redacted].ap-southeast-1.compute.internal 3.10.0-229.el7.x86_64 #1 SMP Thu Jan 29 18:37:38 EST 2015 x86_64 x86_64 x86_64 GNU/Linux
To prepare the Linux instance for kernel-space probing with SystemTap, we will need to install some RPM packages, including kernel-debuginfo and systemtap. We need to ensure that all the kernel packages we will install on the instance have the same version and of the same architecture.
To install a debuginfo package, we’ll enable the “Red Hat Enterprise Linux Server 7 Debug (Debug RPMs)” yum repository that, by default, is disabled.
$ sudo su - # yum repolist disabled Loaded plugins: amazon-id, rhui-lb repo id repo name rhui-REGION-rhel-server-debug-extras/7Server/x86_64 Red Hat Enterprise Linux Server 7 Extra Debug (Debug RPMs) rhui-REGION-rhel-server-debug-optional/7Server/x86_64 Red Hat Enterprise Linux Server 7 Optional Debug (Debug RPMs) rhui-REGION-rhel-server-debug-rh-common/7Server/x86_64 Red Hat Enterprise Linux Server 7 RH Common Debug (Debug RPMs) rhui-REGION-rhel-server-debug-rhscl/7Server/x86_64 Red Hat Enterprise Linux Server 7 RHSCL Debug (Debug RPMs) rhui-REGION-rhel-server-debug-supplementary/7Server/x86_64 Red Hat Enterprise Linux Server 7 Supplementary Debug (Debug RPMs) rhui-REGION-rhel-server-extras/7Server/x86_64 Red Hat Enterprise Linux Server 7 Extra(RPMs) rhui-REGION-rhel-server-optional/7Server/x86_64 Red Hat Enterprise Linux Server 7 Optional (RPMs) rhui-REGION-rhel-server-releases-debug/7Server/x86_64 Red Hat Enterprise Linux Server 7 Debug (Debug RPMs) rhui-REGION-rhel-server-releases-source/7Server/x86_64 Red Hat Enterprise Linux Server 7 (SRPMs) rhui-REGION-rhel-server-rhscl/7Server/x86_64 Red Hat Enterprise Linux Server 7 RHSCL (RPMs) rhui-REGION-rhel-server-source-extras/7Server/x86_64 Red Hat Enterprise Linux Server 7 Extra (SRPMs) rhui-REGION-rhel-server-source-optional/7Server/x86_64 Red Hat Enterprise Linux Server 7 Optional (SRPMs) rhui-REGION-rhel-server-source-rh-common/7Server/x86_64 Red Hat Enterprise Linux Server 7 RH Common (SRPMs) rhui-REGION-rhel-server-source-rhscl/7Server/x86_64 Red Hat Enterprise Linux Server 7 RHSCL (SRPMs) rhui-REGION-rhel-server-source-supplementary/7Server/x86_64 Red Hat Enterprise Linux Server 7 Supplementary (SRPMs) rhui-REGION-rhel-server-supplementary/7Server/x86_64 Red Hat Enterprise Linux Server 7 Supplementary (RPMs) repolist: 0
Now we’ll use yum-config-manager to enable the repo.
# yum-config-manager --enable "Red Hat Enterprise Linux Server 7 Debug (Debug RPMs)"
Now you’ll be able to search for the debuginfo packages.
# yum search kernel | grep ^kernel-debuginfo kernel-debuginfo.x86_64 : Debug information for package kernel kernel-debuginfo-common-x86_64.x86_64 : Kernel source files used by
To ensure that we have the same Linux kernel version, we need to install the kernel and kernel-devel packages along with the kernel-debuginfo packages. If yum returns a different kernel version from the kernel-debuginfo packages, you may need to specify the version manually.
# yum -y install kernel-debuginfo kernel kernel-devel
Depending on your Internet connectivity, it may take some time to download these packages. To give you an idea how large some of these packages are, the kernel-debuginfo-common package is 41MB and the kernel-debuginfo package is 259MB. Fortunately, unless the kernel version and release number change, you only need to install these packages once.
You will need to reboot your machine to apply the new kernel.
# init 6
Once we’re running again, let’s reconnect to the instance.
$ ssh -i "cloudacademy-keypair.pem" ec2-user@[redacted]
Verify that we are using the right kernel version. Notice that the release number is now 229.11.1.el7. Each time Red Hat spins a new kernel, it will update the release number of the package.
$ uname -a Linux ip-[redacted].ap-southeast-1.compute.internal 3.10.0-229.11.1.el7.x86_64 #1 SMP Wed Jul 22 12:06:11 EDT 2015 x86_64 x86_64 x86_64 GNU/Linux
We can use the rpm command to make sure that the version we booted is the same as the debuginfo packages we have installed. This is important.
$ rpm -qa | grep ^kernel kernel-tools-3.10.0-229.el7.x86_64 kernel-3.10.0-229.el7.x86_64 kernel-debuginfo-common-x86_64-3.10.0-229.11.1.el7.x86_64 kernel-3.10.0-229.11.1.el7.x86_64 kernel-devel-3.10.0-229.11.1.el7.x86_64 kernel-tools-libs-3.10.0-229.el7.x86_64 kernel-debuginfo-3.10.0-229.11.1.el7.x86_64
Next, use yum to list all the systemtap-related packages.
# yum list systemtap* Loaded plugins: amazon-id, rhui-lb Available Packages systemtap.x86_64 2.6-10.el7_1 rhui-REGION-rhel-server-releases systemtap-client.x86_64 2.6-10.el7_1 rhui-REGION-rhel-server-releases systemtap-debuginfo.i686 2.6-10.el7_1 rhui-REGION-rhel-server-releases-debug systemtap-debuginfo.x86_64 2.6-10.el7_1 rhui-REGION-rhel-server-releases-debug systemtap-devel.x86_64 2.6-10.el7_1 rhui-REGION-rhel-server-releases systemtap-initscript.x86_64 2.6-10.el7_1 rhui-REGION-rhel-server-releases systemtap-runtime.x86_64 2.6-10.el7_1 rhui-REGION-rhel-server-releases systemtap-sdt-devel.i686 2.6-10.el7_1 rhui-REGION-rhel-server-releases systemtap-sdt-devel.x86_64 2.6-10.el7_1 rhui-REGION-rhel-server-releases systemtap-server.x86_64 2.6-10.el7_1 rhui-REGION-rhel-server-releases
Then install the main systemtap package, and yum will take care of its dependencies.
# yum -y install systemtap
Once all the systemtap-related packages are installed, exit from the root shell to return to the normal user shell.
# logout
Running our first SystemTap script
We have installed all the necessary RPM packages and we are almost ready to run our very first “Hello World” script!
SystemTap uses a simple permission model to grant users’ permissions to use the tool to prevent people from misusing or abusing the tool. Unless you’re a member of the stapusr and stapdev groups, you will not be allowed to build and run SystemTap scripts. Since we are going to write and run our own script, and we hopefully know exactly what we are doing, we will add ourselves to the stapusr and stapdev groups.
$ sudo usermod -G stapusr,stapdev ec2-user $ grep ^stap /etc/group stapusr:x:156:ec2-user stapsys:x:157: stapdev:x:158:ec2-user
You will need to logout and re-login to the instance to reflect that you are now members of these two groups.
$ logout $ ssh -i "cloudacademy-keypair.pem" ec2-user@[redacted]
Type id to see if you are in stapusr and stapdev groups.
$ id uid=1000(ec2-user) gid=1000(ec2-user) groups=1000(ec2-user),156(stapusr),158(stapdev) context=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023
If you are, let’s party on! stap is the SystemTap tool you’ll use to run the script.
There are two ways to write a script. You can either write it in a .stp extension file, or simply specify the script on the command-line with the -e option.
We will opt for the latter since we will use a one-liner script to test if the installation works as expected.
$ stap -e 'probe begin { log("hello world") exit () }'
hello world
Voilà! We have successfully run our first SystemTap script!
In our next article, we will explore some interesting scripts that come with SystemTap.
