2

I've got java logging user activities to Fluentd, Fluentd is then writing these activities into an elasticsearch index.

Every user activity is recorded, some include:

  1. User1 follows user2
  2. User1 likes article1
  3. User1 creates article
  4. User1 searches for tag
  5. user2 signs up
  6. More...

Now for each of these activities I'm storing a user object. For example, a CREATED activity would look like this:

{
activity: "CREATED",
user: {
  userId: X,
  userName: xxxx,
},
{
article: {
  title: XOXO,
  description: "LOTS OF MARKUP",
  date: DATE_CREATED,
  ...more data on article include locations, coords, etc...
  tags: [
     {tagName: "relatedTagToArticle, tagId: 1}
     {tagName: "relatedTag2ToArticle, tagId: 2}
  ],
},
}}

The document could become larger for different activities. But by storing this sort of information I'd be able to select * activities where activities.user = [list of the users followers]and process the results in some sort of algorithm.

Is it fine to keep storing activities like this? Should I avoid this, if so why?

I'm also wondering how I should figure out and store popular tags and searches?

Should I have a program that runs every X minutes and calculates the number of unique SEARCH activities and stores that information in a Redis list?

Edit

One of the main reasons I'd add all this extra information (storing whole objects, such as an article) is so I can query the ES index and get activities from a users followers!

So say I wanted to build a news feed in my app, I could query es with something like this (pseudo code):

{
select all from user_activities
  where {
    activity_type: [LIKE, COMMENT, CREATED, FOLLOW]
    userId: [LIST_OF_FOLLOWING_IDS]
    date > XX AND date < YY
    <SOMETHING HERE TO DETERMINE POPULAR ACTIVITIES>
  }
}

I'm not sure if this is a good approach or whether it will scale or not! Doing it this way, the data stored could become stale, for instance if I displayed a CREATED article from the above query, the article could've been updated! Although I'm not really worried about that just yet.

  • Why not use Google Analytics? – Prasanna K Rao Sep 9 '16 at 6:21
  • @PrasannaKRao why would I use Google analytics, that just makes me have to configure another service right? I'd rather just use es for storing user activities and somehow count the most popular tags, locations, searches, and etc! – James111 Sep 9 '16 at 6:23
  • Ok. It's best if you store data in es that is required for search only. I don't see you using that data for search. One common technique is to boost the quality of a document itself, based on such user activities. if you store such info in GA (also), building reports like top 10 searches, top 10 countries/departments from which the system is used, is very easy. If you do want to store these in es, why store the whole doc? – Prasanna K Rao Sep 9 '16 at 6:40
  • Say I wanted to curate a feed based on a users followers, I could simply query the user activity index, which would return data that could be directly returned to the app and diaplyed. I'm not sure if that's a common practice or not? If I didn't do this I'd store ids and then query mysql (this would probably take a lot longer). @PrasannaKRao – James111 Sep 9 '16 at 6:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.