Rolling updates, zero downtime, clustering and load balancing | IFS Community
Solved

Rolling updates, zero downtime, clustering and load balancing

  • 16 January 2020
  • 3 replies
  • 295 views

Userlevel 4
Badge +6

Hi,

I would like to know if Enterprise Explorer would handle rolling updates in this scenario:

  • Cluster of 2+ nodes
  • NGINX load balancer

 

The process for applying updates would be:

  1. Take down 1 node
    • Does Enterprise Explorer switch over to a new node seamlessly without crashing and kicking the user out?
  2. Apply the client delivery to that node
  3. Bring the node back up
  4. Repeat the process for all nodes…
  5. Profit?

Is this scenario possible without EE crashing?

 

NOTE: We only deliver client updates using the installer. Database changes are done manually to prevent downtime. This also cuts the delivery process to a couple of minutes, instead of hours.

 

icon

Best answer by MikeArbon 18 February 2020, 09:49

View original

3 replies

Userlevel 5
Badge +9

Hi,

IFS recommendation is to install the update using the installer with the instructions provided in the update.

Becuase IFS updates are not tested in a way like you have mentioned.

Further, apart from the database and client part, there is another part in an update (application server changes).

/Savinda

 

Userlevel 5
Badge +10

Hi

It is a good question worth asking, but i wonder if this is a no.

Which version of IFS are you on? My answer below is based on IFSv9.

IFS of course recommend that there is no activity on the system during the application of a release. For me this is the main problem, as if you run into issues, you will be exposed to the unknown and IFS support may be unable to assist, resulting in serious downtime during a restore.

That aside, batch queues are supposed to be disabled (affecting all nodes), a backup of IFS home is required (I suspect the system needs to be fully shutdown for a clean backup). A release for a clustered environment requires the certificate being signed before creation of the cluster zip files, could this present an issue to those on the running nodes? Stand alone items such as Print agents, Crystal reports, Demand Planner all may be affected by a release….

An interesting query, which would be very useful if there was an IFS supported procedure. You can probably tell I am a risk adverse System Admin….

Regards

Mike

Userlevel 4
Badge +6

Hi

It is a good question worth asking, but i wonder if this is a no.

Which version of IFS are you on? My answer below is based on IFSv9.

IFS of course recommend that there is no activity on the system during the application of a release. For me this is the main problem, as if you run into issues, you will be exposed to the unknown and IFS support may be unable to assist, resulting in serious downtime during a restore.

That aside, batch queues are supposed to be disabled (affecting all nodes), a backup of IFS home is required (I suspect the system needs to be fully shutdown for a clean backup). A release for a clustered environment requires the certificate being signed before creation of the cluster zip files, could this present an issue to those on the running nodes? Stand alone items such as Print agents, Crystal reports, Demand Planner all may be affected by a release….

An interesting query, which would be very useful if there was an IFS supported procedure. You can probably tell I am a risk adverse System Admin….

Regards

Mike

We are on IFS 10 UPD5, soon on UPD7.

I see, these are indeed serious concerns.

I guess we’ll just wait until Enterprise Explorer gets totally replaced by Aurena, just like the web client in Business Central did with the RTC client recently.

 

Thanks for all the valuable insight. :blush:

Reply