t2.micro instances in ECS guide are too anaemic

[rade]

The ECS guide/script starts t2.micro instances. These instance types operate with CPU credits. Once the credits are exhausted, the machine grinds to a halt until sufficient time passes for credits to be replenished. This makes these instances types a poor choice for experimentation, since the dramatic performance degradation can be mistaken for a problem with weave scope/net.

cc @2opremio

[2opremio]

Yep, noticed this in weaveworks/scope#897 but AFAIK only t2.micro instances are covered by the free tier. We could add a warning mentioning that a beefier type of instance will be needed for setups running for more than a couple of hours.

[rade]

A warning in the setup script? Won't that get drowned by the general noise?

We may also want to add some pointers for finding out whether VMs are being throttled.

[2opremio]

A warning in the setup script? Won't that get drowned by the general noise?

I meant the guide but we could do both. In fact we will be able to show it more explicitly when we replace the script with Cloudformation.

We may also want to add some pointers for finding out whether VMs are being throttled.

I don't myself know how to do this but I will look into it.

[errordeveloper]

I don't see a good reason to work around this, I think we should just use m3 or other type of instances of substantial size.

[2opremio]

I don't see a good reason to work around this, I think we should just use m3 or other type of instances of substantial size.

Only t2.micro instances are covered by the free tier. Remember it's a guide which people probably will use for testing, we don't want them to spend money if not absolutely necessary.

[errordeveloper]

Sure, but may be that is less important whether they will spend a few
dollars or not... they will have card on file at this point anyway.

On Wed, 8 Jun 2016, 09:54 Alfonso Acosta, [email protected] wrote:

I don't see a good reason to work around this, I think we should just use
m3 or other type of instances of substantial size.

Only t2.micro instances are covered by the free tier. Remember it's a
guide which people probably will use for testing, we don't want them to
spend money if necessary.

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#181 (comment),
or mute the thread
https://github.com/notifications/unsubscribe/AAPWS8eL3ZDq8v0GtidS6hJwPrvEAs0cks5qJoMrgaJpZM4IvJ8U
.

[2opremio]

they will have card on file at this point anyway.

I don't think we should be assuming this

[2opremio]

Another use run into this https://weaveworks.slack.com/archives/weave-users/p1466154335000239

However, he is using t2.micro instances in production, using larger instances apparently isn't an option for him.

[errordeveloper]

When I set up my account a few years ago, I remember I couldn't do much
before I added a card.

On Fri, 17 Jun 2016, 10:09 Alfonso Acosta, [email protected] wrote:

they will have card on file at this point anyway.

I don't think we should be assuming this

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#181 (comment),
or mute the thread
https://github.com/notifications/unsubscribe/AAPWS0ifyXMt36Nm1eJMm5qiWmNVVm9uks5qMmRYgaJpZM4IvJ8U
.

[2opremio]

Yes, I meant we shouldn't be assuming that people can spend money away even if the card is added.

[errordeveloper]

Yes, I meant we shouldn't be assuming that people can spend money away even if the card is added.

Ok, that's true.

[errordeveloper]

In any case, we should start working on #175 and figure out some nice UX. Current page is a bunch of screenshot, and it can be turned into a much more enhanced experience, where people will learn information like what's being discussed here. We can also consider merging it with the current plain text guide or linking in some cleaver way.

[rade]

Perhaps we should add a note somewhere explaining the symptoms and linking to instructions for discovering when instances have been throttled.