Is there a way of referencing an entity at a point in time? I want a record to reference the value of an exchange rate at the time it was created. I can create a copy of the various bits of info, but I just wondered if I could instead simply reference the entity id and db and have that realised in the query

@p14n You could store the transaction id or the date of the transaction of the point in time you're interested in.

Yup, I was thinking along those lines. Means that I'd need to perform 2 queries (one to get the tid and one to get the value)

Thanks

You will have to query with a database value at the time you're interested in anyway I think. But if this value is going to be queried a lot I would store it separately in the database instead of having to query the old database value every time.

I think you're right. I was just checking as referencing another entity at a point in time seemed like a 'natural' fit for datomic

if you need to store this reference in datomic itself, you can create a ref attribute that points to the tx

e.g. {:snapshot/entity eid :snapshot/time tx-eid}

Does anybody know the current query group instance size options?

I cant find them documented anywhere (I’m sure they’re subject to change, which may be why)

@joe They are in the QG CF template. You can choose from the sizes there. Let me pull it up.

79     "InstanceType": {
  80       "Description": "Cluster node instance type",
  81       "Default": "t2.medium",
  82       "AllowedValues": [
  83         "t2.medium",
  84         "m5.large",
  85         "i3.large",
  86         "i3.xlarge"
  

Thanks @jaret

@jaret Follow up question. If I were interested in using ~10 GB of that tasty 500 GB NVMe SSD on the production i3.large instances (or from an m5.large query group) are there any datomic cloud issues around that? Can I write to a directory on disk from an ion? The use case is a Lucene index where documents are filtered by data coming back from datomic. I know this is a potential can of worms if I don’t keep the index size managed well, but lets assume this can be done. Asking for a friend who is estremely dissatisfied with ElasticSearch…

Said friend is also happy to do the index management off on a query group, so as to not affect the primary groups what so ever.

@joe You should be able to do something like that. We make minimal use of disk, and have done nothing to get in the way of using the boxes (that we know of). The important caveat here is we have not tested/vetted writing to a directory from an Ion or implementing Lucene etc. So I cannot officially recommend the approach and it remains untested/unsupported. All that being said, you should try/test it and let us know how it goes. 🙂

My “friend” is extremely excited