I've setup a summary index that works great. I usually use it like this:
index=summary search_name="Z - Top Domain - 15 minutes" | top 50 http_domain
When I run this query, it gives me three fields - the http_domain, the "count" which is the # of hits of that domain, and the percent of total. I'd like to graph the sum(count) by http_domain. I naively tried this, and of course failed hard. Attempts to use bucket have met with blank stares from the search engine as well.
It's clear the data is in there - if I run the query for 1 hour, I get 1 hour of results. Ideally then I could bucket the results each hour or equivalent.
How are you populating the summary index? If you're using "... | sitop http_domain" then the "... | top" is really the only valid thing you can do.
On the other hand, if you store the summary manually, say "... | stats count by http_domain", then you should be able to compute "... | timechart sum(count) by http_domain".
Calculating "... | top 50 http_domain" is a bit harder, say "... | stats sum(count) as count by http_domain | sort - count | head 50" and is even harder if you want percents, where you'll have to add "... | eventstats sum(count) as sum_count | eval percent = count / sum_count | fields - sum_count | ..." between the stats and the sort.