Freeze on ensure-index

[pholz]

My application occasionally freezes when calling ensure-index on a db which seems to be connected. In the code below, I call connect and then initialize-indexes, and "initializing indexes" is the last output I see. The code never reaches the next info output in the function that called initialize-indexes. I cannot reproduce this in the REPL, or locally, it happens once every few weeks with the uberjar on the production machine. There is no error, the program does not exit, it just freezes completely and needs to be restarted.

monger 3.1.0, mongodb 5 in a docker container.

What is the best way to debug this?

(defn connect []
  (reset! gldb (let [conn (mg/connect-with-credentials
                           "127.0.0.1"
                           (mcr/create "mySrv" "admin" (.toCharArray "myPass")))]
                 (info "connecting to db: " conn)
                 (mg/get-db conn "myDb"))))

(defn initialize-indexes []
  (info "initalizing indexes")
  (mc/ensure-index @gldb "weather-monthly" (array-map :time 1 :station 1) {:unique true})
  (mc/ensure-index @gldb "weather-10min" (array-map :time 1 :station 1) {:unique true})
  (mc/ensure-index @gldb "airq" (array-map :time 1 :station 1) {:unique true})

  (mc/ensure-index @gldb "grid-wind-s" (array-map :time 1) {:unique true})
  (mc/ensure-index @gldb "grid-wind-m" (array-map :time 1) {:unique true})
  (mc/ensure-index @gldb "grid-wind-l" (array-map :time 1) {:unique true}))

[michaelklishin]

Server logs or a traffic capture. An index can take a while to build.

[pholz]

thank you. I looked at the server logs. When I initally call connect, everything is ok. Then, on the first call to ensure-index the mongo logs show that a second connection is established. Is this normal? In most cases, successive calls to ensure-index complete (and in any case they do not cause further connections to be established). Once in a while, the connect calls succeeds but then the connection attempt from the first call to ensure-index seems to fail (and does not recover even after hours). Is it wrong to save the DB connection in the atom and then use that?