Hi,

I enable impala admission control with a cluster which has two impala
coordinators and six impala executors.

Max Memory Multiple : 8GB
Minimum Query Memory Limit : 1GB
Maximum Query Memory Limit : 5GB

impalad version 3.4.0

There were some queries which were admitted to the cluster and allocated
memory within the memory range by impala admission control and gave the
below exception while executing the queries.

Status: Memory limit exceeded: Error occurred on backend
Memory left in process limit: 22 .63 GB Memory left in query limit :
-352.28 KB
Query (7c4b6e3d86b1cb83 :56c29eef00000000) : memory limit exceeded.
Limit=1.00 GB
Reservation=818 .00 MB ReservationLimit=819.40 MB OtherMemory=206.59 MB
Total 1.00 GB
Peak=1.GB Fragment 7c4b6e3d86b1cb83: Reservation=818.00 MB
OtherMemory=206.59 MB Total-1.00GB Peak=1.00 GB SORT NODE (id=6) :
Total=197.51 MB
Peak=197.51 MB SELECT NODE (id=5) : Total=20.00 KB Peak=8.02 MB Exprs:
Total=4.00 KB
Peak=4.00 KB ANALYTIC_EVAL_NODE (id=4) : Reservation=4.00 MB
OtherMemory=9.01 MB
Total=13.01 MB Peak=14.04 MB Exprs: Total=4.00 KB Peak=4.00 KB SORT NODE
(id=3) :
Reservation=814.00 MB OtherMemory=16.00 KB Total=814.02 MB Peak=814.02 MB
EXCHANGE_NODE
(id=7) : OtherMemory=0 Total=0 Peak=21.28 MB KrpcDeferredRpcs: Total=0
Peak=26.29 KB KrpcDataStreamSender (dst_id=8) : Total=168 .00 B Peak=168.00
B CodeGen:
Total=28.24 KB Peak=5.23 MB Fragment 7c4b6e3d86b1cb83:56c29eef00000005:
Reservation=0
OtherMemory=0 Total=0 Peak=20.90 MB UNION NODE (id=0) : Total=0 Peak=1.39 MB
KUDU_SCAN_NODE (id=1) : Total=0 Peak=18.08 MB HDFS_SCAN_NODE (id=2) :
Reservation=0
ReservationLimit=0 OtherMemory=0 Total=0 Peak=4.00 KB KrpcDataStreamSender
(dst_id=7) :
Total=0 Peak=168.98 KB CodeGen: Total=0 Peak=4.58 MB


Can you assist me find a solution for this issue ?

Thanks & Regards
Donishka

Reply via email to