Currently working on an e-commerce solution, we are using Elasticsearch 6.x to enable full-text search on products.
We have 88 indices, each one having a primary and a replica. For each index, primary.store.size is around 50mb. We are often regenerating those indices through bulk indexation with fixed _id. Because of that, translog is growing.
In its default configuration, translog retention size is 512mb and retention age is 12h. According to documentation, translog is now kept for a longer time to enable a faster shard recovery mechanism. In my scenario, translog is taking much more space than indices.
Given the size of our indices which is pretty small compared to the translog size, could I tweak safely the translog retention age or retention size? Or would I better increase my storage space to take into account the default translog properties ?
PS: I have posted the same question on stackoverflow at this link: https://stackoverflow.com/questions/56108350/how-to-tune-translog-retention-for-small-indices. Should I remove it to keep it only here?