# Startup time impact of mender

**URL:** <https://hub.mender.io/t/startup-time-impact-of-mender/2190>\
**Category:** General Discussions\
**Tags:** yocto, nxp, warrior, imx6ul-evk, analytics\
**Created:** [July 13, 2020, 10:57am UTC](https://hub.mender.io/t/startup-time-impact-of-mender/2190 "2020-07-13T10:57:00Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![danie](https://yyz2.discourse-cdn.com/flex036/user_avatar/hub.mender.io/danie/32/620_2.png) [@danie](https://hub.mender.io/u/danie)\
**Post date:** [July 13, 2020, 10:57am UTC](https://hub.mender.io/t/startup-time-impact-of-mender/2190/1 "2020-07-13T10:57:00Z")

</div>

Hi,

I’ve successfully integrated mender into NXP’s iMX 6UL by Yocto Warrior. And I’m analysing the startup time of the board with ‘`systemd-analyze`’.

My concern is, Mender is not shown in the services list obtained by running ‘`systemd-analyze blame`’ but mender is running as daemon (confirmed by running `pidof`).

So,

1. Why mender daemon is not listing in `systemd-analyze`?
2. how can I measure the startup impact of mender? Are there any particular method or do you have any benchmarks?

Thanks in advance,  
Daniel.

---

<div class="post-metadata">

**Author:** ![dellgreen](https://yyz2.discourse-cdn.com/flex036/user_avatar/hub.mender.io/dellgreen/32/85_2.png) [@dellgreen](https://hub.mender.io/u/dellgreen)\
**Post date:** [July 13, 2020, 11:33am UTC](https://hub.mender.io/t/startup-time-impact-of-mender/2190/2 "2020-07-13T11:33:33Z")

</div>

I think the blame command lists only running service units, which implies that the mender-client service unit file is considered as completed. If you do “systemd-analyse plot \> mender.svg” it should be listed in there.

I’m not overly familiar with the mender client binary itself but:  
given that in ExecStart= in the service unit file the mender binary is daemonizing itself I would expect the unit type to be something like “forking” so that the child process is monitored by systemd and the parent process can exit.

You could try changing the unit file type to ‘forking’ as a proof of concept to see if its a bug, or whether there is a specific reason that the type is ‘idle’

---

<div class="post-metadata">

**Author:** ![kacf](https://yyz2.discourse-cdn.com/flex036/user_avatar/hub.mender.io/kacf/32/146_2.png) [@kacf](https://hub.mender.io/u/kacf)\
**Post date:** [July 15, 2020, 6:09am UTC](https://hub.mender.io/t/startup-time-impact-of-mender/2190/3 "2020-07-15T06:09:56Z")

</div>

`forking` will not work as the Mender daemon does not actually fork. `simple` is the recommended option for services that neither fork nor provide startup notification, and `idle`, which Mender uses, is just a more relaxed version of `simple`. AFAIK it does not provide any information about startup time, which would require either forking or notification support.

---

<div class="post-metadata">

**Author:** ![danie](https://yyz2.discourse-cdn.com/flex036/user_avatar/hub.mender.io/danie/32/620_2.png) [@danie](https://hub.mender.io/u/danie)\
**Post date:** [July 15, 2020, 7:25am UTC](https://hub.mender.io/t/startup-time-impact-of-mender/2190/4 "2020-07-15T07:25:32Z")

</div>

> [@danie](#):
>
> How can I measure the startup impact of mender? Are there any particular method or do you have any benchmarks?

Does somebody have any idea on this?