Query with "failOnClusterChange" breaks command queue

Queries with “failOnClusterChange” and secondary index filter are instantly timing out once they are getting queued. The test passes correctly if ‘maxCommandsInProcess’ setting is high enough or if no secondary index is used.

Test to reproduce:

it('queue with failOnClusterChange', async (done) => {
  Aerospike.setupGlobalCommandQueue({ maxCommandsInProcess: 5, maxCommandsInQueue: 50 })
  client = await Aerospike.connect({
    hosts: '127.0.0.1:3000',
    policies: {
      query: new Aerospike.QueryPolicy({ totalTimeout: 10000, failOnClusterChange: true })
    }
  })

  try {
    await client.createIntegerIndex({
      ns: 'test',
      set: 'test',
      bin: 'i',
      index: 'test'
    })
  } catch (e) {
    //
  }

  const puts = Array.from({ length: 5 }, (_, i) => client.put(new Aerospike.Key('ssd-store', 'test', i), { i }))

  await Promise.all(puts)

  const queries = Array.from({ length: 5 }, async (_, i) => {
    const query = client.query('ssd-store', 'test')

    // does not work with sencondary index
    query.where(filter.equal('i', i))
    // works with predexp
    // query.where([predexp.integerBin('i'), predexp.integerValue(5), predexp.integerGreaterEq()])
    await query.results()
  })

  await Promise.all(queries)
  client.close()
  done()
})

Result: AerospikeError: Delay queue timeout

[jhecking]

I’ve just released v3.16.2 to npmjs.com.

[jhecking]

A new C client version 4.6.19 is ready, which includes a fix for this issue. I’m planning to release Node.js client version 3.16.2 tomorrow.