none
RDS 2012: best practises daily reboot? RRS feed

  • Question

  • Hi,

    Simple question: what would be best practises to reboot a RDS 2012 server (in a 4 server farm with broker) on which lots of applications and users are. I guess a daily reboot would be best, right?

    Please advise.
    J


    Jan Hoedt

    Friday, September 23, 2016 8:22 AM

Answers

  • Hi,

    As a rule, you only restart a server when you have a specific reason or reason(s) to.  For example, some people make a schedule and install windows updates once per month, after testing them in staging environment.

    In highly-controlled environments, I have seen heavily used RDSH servers running constantly for six months or more.  Now, they are not typical, since in most cases it is best to install security updates regularly, but I wanted to let you know it is certainly possible to do.

    I disagree with the premise that rebooting is a simple way to proactively prevent issues.  What problem do you think needs to be or is solved by restarting regularly?

    If a specific issue (such as memory leak) is identified and cannot be fixed with an update or configuration change, in many cases it can be fixed by restarting the leaky service(s).  If in this example it is not possible to restart the service, then perhaps a regular restart of the entire server would be the best remedy.  Cases like I describe in this example are the rare exception.

    One thing to note is environments are vastly different and thus a "good average" for restarting servers wouldn't have much meaning.  By different I mean, number of applications, number of users, RAM/CPU/IO capacity, specific type of third-party applications (e.g. whether or not they have service(s) or filter drivers), printer drivers, security requirements, etc.

    As far as re-provisioning all the RDSH servers on a regular basis, that is different than restarts.  One reason is, when you need to apply updates to a group of RDSH servers, it is possible for one or more updates to fail to apply to one or more RDSH servers, causing them to be different.  Applying the updates to a master image template and using that as the basis for all the RDSH servers promotes uniformity as you mentioned.

    One thing I will say is, there should not be a reason to rebuild all the RDSH servers every month "just because" without there being new updates/changes or some identified issue that is being corrected by the re-provision.

    Many years ago it was fairly common to set up terminal servers to reboot on a regular schedule, however, there has not been a need to do that anymore in my experience (except in vary rare case).

    -TP

    • Proposed as answer by Amy Wang_Moderator Tuesday, September 27, 2016 1:36 AM
    • Marked as answer by janhoedt Thursday, September 29, 2016 1:08 PM
    Monday, September 26, 2016 4:25 PM
    Moderator

All replies

  • Hi,

    There is no need to restart servers daily.

    If at some point in the future you identify a specific issue and after troubleshooting you determine the only/best way to mitigate the issue is regular restarts then it is something to consider.

    -TP

    • Proposed as answer by Amy Wang_Moderator Monday, September 26, 2016 7:44 AM
    • Unproposed as answer by janhoedt Monday, September 26, 2016 9:29 AM
    Friday, September 23, 2016 12:34 PM
    Moderator
  • Don't agree. It doesn't make sense to leave servers running for a year even if they have no issues. However daily is maybe also not needed. Therefore I'm looking for a good average. Yes, it depends on number of users and applications but still then there should be a good average.

    Rebooting is a simple way to proactively prevent issues. Specifically because you can always should find a small maintenance window without user impact. 

    Some companies even restage rds/citrix servers each month specifically for that reason: stability/uniformity in the environment.


    Jan Hoedt

    Monday, September 26, 2016 9:29 AM
  • Hi,

    As a rule, you only restart a server when you have a specific reason or reason(s) to.  For example, some people make a schedule and install windows updates once per month, after testing them in staging environment.

    In highly-controlled environments, I have seen heavily used RDSH servers running constantly for six months or more.  Now, they are not typical, since in most cases it is best to install security updates regularly, but I wanted to let you know it is certainly possible to do.

    I disagree with the premise that rebooting is a simple way to proactively prevent issues.  What problem do you think needs to be or is solved by restarting regularly?

    If a specific issue (such as memory leak) is identified and cannot be fixed with an update or configuration change, in many cases it can be fixed by restarting the leaky service(s).  If in this example it is not possible to restart the service, then perhaps a regular restart of the entire server would be the best remedy.  Cases like I describe in this example are the rare exception.

    One thing to note is environments are vastly different and thus a "good average" for restarting servers wouldn't have much meaning.  By different I mean, number of applications, number of users, RAM/CPU/IO capacity, specific type of third-party applications (e.g. whether or not they have service(s) or filter drivers), printer drivers, security requirements, etc.

    As far as re-provisioning all the RDSH servers on a regular basis, that is different than restarts.  One reason is, when you need to apply updates to a group of RDSH servers, it is possible for one or more updates to fail to apply to one or more RDSH servers, causing them to be different.  Applying the updates to a master image template and using that as the basis for all the RDSH servers promotes uniformity as you mentioned.

    One thing I will say is, there should not be a reason to rebuild all the RDSH servers every month "just because" without there being new updates/changes or some identified issue that is being corrected by the re-provision.

    Many years ago it was fairly common to set up terminal servers to reboot on a regular schedule, however, there has not been a need to do that anymore in my experience (except in vary rare case).

    -TP

    • Proposed as answer by Amy Wang_Moderator Tuesday, September 27, 2016 1:36 AM
    • Marked as answer by janhoedt Thursday, September 29, 2016 1:08 PM
    Monday, September 26, 2016 4:25 PM
    Moderator
  • It is disappointing that you did not actually attempt to answer this persons question. Is there an best practice for scheduling a regular (weekly, monthly) reboot.

    I personally love servers that run for a year without issue - but RDS application servers like windows desktop clients run 3rd party software applications. Occasionally there are required reboots to solve a problem.

    A good answer to what is the best practice for RDS 2016 server scheduled reboots would be helpful.

    Thanks in advance,

    Bruce

    Thursday, August 22, 2019 2:48 PM