Proposal, change default tsdb retention to be much bigger (or smaller)

[baryluk]

Hi,

I had my prometheus running for few months, but I just noticed that (after upgrading to 2.6~rc0), that I can't see any data older than few days. I have few small instances of prometheus that were configured quickly without much rules or configs, other than simply adding few static targets (monitoring single target, or just few small servers / network appliances for getting some data in).

Reading https://prometheus.io/docs/prometheus/latest/storage/ , it appears that default retention (--storage.tsdb.retention is just 15 days.

Obviously for serious production I would configure it manually, but in many cases where I just startup prometheus for a one server, or run it for development, I would prefer to have default retention be much longer out of the box, to minimize configuration.

I can't fine any reasoning behind setting default of 15d.

I would prefer much more running out of disk space, and getting some alert, than loosing data so early by default.

I would propose to change default to either significantly less (to force custom retention, and be easy to spot quickly), or to something usable by most without any tweaks. That is either 1d, or 1y.

[brian-brazil]

We had discussed and agreed on this for 2.0 but noone made the change. As this is a breaking change (some users may not have the disk for this) we have to wait for 3.0, whenever that is.

…

On Sun 9 Dec 2018, 03:24 Witold Baryluk ***@***.*** wrote: Hi, I had my prometheus running for few months, but I just noticed that (after upgrading to 2.6~rc0), that I can't see any data older than few days. I have few small instances of prometheus that were configured quickly without much rules or configs, other than simply adding few static targets (monitoring single target, or just few small servers / network appliances for getting some data in). Reading https://prometheus.io/docs/prometheus/latest/storage/ , it appears that default retention (--storage.tsdb.retention is just 15 days. Obviously for serious production I would configure it manually, but in many cases where I just startup prometheus for a one server, or run it for development, I would prefer to have default retention be much longer out of the box, to minimize configuration. I can't fine any reasoning behind setting default of 15d. I would prefer much more running out of disk space, and getting some alert, than loosing data so early by default. I would propose to change default to either significantly less (to force custom retention, and be easy to spot quickly), or to something usable by most without any tweaks. That is either 1d, or 1y. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#4981>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGyTdk3zGVUvNhvXh-Jr9DfFNxUn7aM_ks5u3HRZgaJpZM4ZJ6mi> .