-
Notifications
You must be signed in to change notification settings - Fork 224
aiming to close spot-23 #24
base: master
Are you sure you want to change the base?
Conversation
aims to close the #spot-23 jira with pr #24 |
Spot-23: https://issues.apache.org/jira/browse/SPOT-23 Hey, @natedogs911 would you paste the discussion part in the JIRA issue? |
I edited my comments, nfdump is the known issue, the rest are things that i want to improve/add |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If there are no particular reasons to keep separated scripts in ML and Ingest, I would put the installation all together in a single script.
If the scripts get too long, then I would recommend writing different pieces but then have only one one script to call each functionality ( build, post build, etc.).
spot-ingest/install.sh
Outdated
pip install -r requirements.txt | ||
fi | ||
|
||
log_cmd "dependencies satisfied, please run ./build.sh to complete setup" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
spot-ingest install.sh is asking to run build.sh but there is no spot-ingest/build.sh, is this message something should be here or build.sh is missing?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Originally i did have a build.sh, but as i refactored things it became redundant.
I simply forgot to update this message.
spot-ml/README.md
Outdated
|
||
1. run `sudo ./install.sh` | ||
2. run `./build.sh` | ||
3. run `sudo ./post_build.sh` |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I vote to put these three together in a single script and just print the "state" of the installation like "SBT installation, SBT assembly and JAR installation".
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The reason for separate scripts is because you need root for the installation and post build steps, if the consensus is that root should be used for all of these steps then i agree.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What about using sudo -H -E? I'm not even sure that's the syntax but I remember using something like that to elevate permissions but still keep using current user.
## Install | ||
|
||
1. run `sudo ./install.sh` | ||
2. run `hdfs_setup.sh` |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Wondering if it's possible to automatize the step of moving spot.conf
to /etc/ folder in Spot Installation Documentation, step 3.4.
Also, about your point for discussion, this PR assumes then that everything including spot-ml, spot-ingest and spot-oa is going to be landed in a single node, am I correct?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I agree, spot.conf could be added very easily.
However i was thinking i might do move that to /opt/spot/spot.conf instead. this requires changing each components file path which should be fairly simple.
In regards to the discussion,
It would be simple to have a install-all.sh
script that ran everything,
#!/bin/bash
source ./install.sh
source ../spot-ingest/install.sh
source ../spot-ml/install.sh
source ../spot-oa/install.sh
and then a user can make the choice if they are landing individual components elsewhere.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
In case users want to land in different nodes, then they would need to run separate scripts but if they do in a single node, then just run install-all.sh, is that what you are saying?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
that's correct
Csv removal final
Ready for review and merge... |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I have a few questions:
- About documentation changes, are you talking about Spot web page changes? Or more README changes?
- When you say we need error handling, are you saying we need a "standard" way to print errors? Or you are talking about actually handling errors? Are any errors being handled in this version?
- I don't see spot.conf
scp
, are users still required to copy manually? It would be really great if we can make this automatic. - I don't see install-all.sh either, so the decision was not to have a script like that?
- Finally, could you please write a quick description of how the installation process is going to be after this PR?
Thanks.
+1 |
RE: @rabarona
and change the spot.conf file to spot.conf.template
a. from the ./spot-setup run b. run c. per ./spot-setup/README.md run d. from ./spot-ingest run e. while in ./spot-ml run f. while in ./spot-oa run If you want to run a particular component such as spot-ingest on another node, simply copy /etc/spot.conf to that node along with the spot-ingest folder and run the ./install.sh for that component. |
how about to simplify the 6 step process outlined above we can add an --install-all option to the spot-setup/install.sh script which executes the script for each component? |
QA and Review and advise on any changes please,
This will require some doc changes.
known issues:
TBD:
Discussion:
should the assumption be made that all components are being landed on the same server?
This would greatly simplify things as in spot-setup you could just run a script to source each components scripts. feedback would be great.
EDIT: really there is only the one known issue, nfdump is in the spot-nfdump repo that i'm pulling in and must be fixed there.