EPIC: Rebuild / Upgraded AWS EC2's

[RobHanna-NOAA]

Note: Now part of 1377 EPIC: FIM Sys Admin Tasks (and a few related FIM tasks)

There are now 7 fixes and updates required or suggested for EC2's. Not all are mandatory but all have value for various reasons. If we do drop back to small ECS'2, we can fix the base model and it will be auto in place for all team EC2's. If we do not replace everyone's, we will need scripts to figure this out.

Note: many of these tasks are tightly interconnected.

Many of these tasks a very large amount of time. This card arguable could be an epic card.

Many of these will also required updating the AWS FIM dev guide as well. All of these task have had significant tests and research done on them but are not complete. Those notes are very extensive and not recorded here. I (Rob) have those notes which we can figure out if/where to put those if we want too.

Tasks:
1) Fix the various problems related to security and updates. We are having a lot of trouble with auto updates and what are updated in those updates. Some of the updates run on some machines have already recorded system errors that show up on all start ups. It would be nice to move Ubuntu from 22.04 for 24.04, but the upgrades are well noted as being unstable. We may consider starting with new Ubuntu 24.04 machines as a possible answer to some of the other problems and fixes below. This is one of our single biggest issues.

Building entirely brand new Ubuntu 24.04 images from scratch is a major task but HV likely has some templates we can start with. I have a large list of requirements and adjustments that will be needed if we start over. We can also decide if we even want to go with Ubuntu this time as it is not necessarily the only options. Choices of desktop software are also key factors.

These update problems are also showing various "segmentation faults" which is a memory usage problem at the root Ubuntu level.

2) Segmentation Faults at the Ubuntu OS Level. Depend on our choices from Option 1 will dictate what we do here. This has now been split to it's own issue 1376 due to urgency

3) Tagging: As new images are created, we need to add to the "first time login" script to ensure tags are added to new EC2's as being created. (easy and quick)

4) See if we can fully remove LibreOffice, then install just LibreCalc. Why? We are seeing some strange errors when the EC2 attempts to auto update it. If this is even possible, it is less software the system it attempting to update.

5) FIM EC2: can we get Libre to quit giving an error when opening csv's on EFS?

6) Remove Notepad++ as it is not used. Just a cleanup issue and less to let the system try to do updates on.

7) Look into RDP speeds. We have a problem fix but does require a script to have people run on their machine and also the "first login" script.

8) QGIS / Parquet. Our current image will not let us open Parquet files in QGIS. The reasons are most related to QGIS version releases for Ubuntu specifically. This is a deceptively complex story. If we can move up to Ubuntu 24.04 (not by upgrading our 22.04), this will likely give us the dependencies to let Parquet work in QGIS. Note.. the problem is not so much QGIS needing upgrades but mostly the OS version of GDAL and it's version compatibilities.

• Update: Dec 16, 2024: Our docker builds rely on our current version of GDAL and python version on our EC2's. This has to stay compatible with our external users ability to it is well. Upgrading both creates problems for the HAND product. Lots of research required and maybe should be it's own card.

This is proving to be a very complex fix, but again.. depending on if we build new images from scratch.