# High Heap usage after a while

**URL:** <https://community.cratedb.com/t/high-heap-usage-after-a-while/897>\
**Category:** CrateDB\
**Created:** [November 18, 2021, 3:01pm UTC](https://community.cratedb.com/t/high-heap-usage-after-a-while/897 "2021-11-18T15:01:55Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![fox\_pluto](https://sea2.discourse-cdn.com/flex020/user_avatar/community.cratedb.com/fox_pluto/32/417_2.png) [@fox\_pluto](https://community.cratedb.com/u/fox_pluto)\
**Post date:** [November 18, 2021, 3:01pm UTC](https://community.cratedb.com/t/high-heap-usage-after-a-while/897/1 "2021-11-18T15:01:55Z")

</div>

Hi, I have build a cluster with 3 nodes of 8 cores and 8 GB each, the `CRATE_HEAP_SIZE` is set to 2G.

I have used the cluster for some days for ingesting end retrieving jsons, not the cluster report:

```auto
SQL Error [XX000]: ERROR: out_of_memory_error: Java heap space

```

the Heap from the GUI is actually full, even if I am not using the cluster any more.

What this heap is used for?  
How could I prevent the situation? the 2 GB of heap is too low?  
Is there a way to empty the Heap ?

Regards,  
S.

---

<div class="post-metadata">

**Author:** ![proddata](https://sea2.discourse-cdn.com/flex020/user_avatar/community.cratedb.com/proddata/32/1379_2.png) [@proddata](https://community.cratedb.com/u/proddata)\
**Post date:** [November 18, 2021, 3:11pm UTC](https://community.cratedb.com/t/high-heap-usage-after-a-while/897/2 "2021-11-18T15:11:02Z")

</div>

> What this heap is used for?

some state data for tables, intermediate results, and various other objects

> How could I prevent the situation? the 2 GB of heap is too low?

Depends. With the right setup you could store and index quite some data

> Is there a way to empty the Heap ?

This is done automatically by the JVM GC, as long as unused Objects can be deleted

* * *

This seems a bit like an overload situation

Could you give some context about how much data you ingested, what the table schema looks like.  
How many shards did you create?

You could increase the heap to ~50% of the node memory

---

<div class="post-metadata">

**Author:** ![fox\_pluto](https://sea2.discourse-cdn.com/flex020/user_avatar/community.cratedb.com/fox_pluto/32/417_2.png) [@fox\_pluto](https://community.cratedb.com/u/fox_pluto)\
**Post date:** [November 18, 2021, 3:24pm UTC](https://community.cratedb.com/t/high-heap-usage-after-a-while/897/3 "2021-11-18T15:24:43Z")

</div>

Hi,

yes of course, I have 3 nodes each with 3 x 1TB SSD disk.  
I have 3 tables created in this way:

```auto
create table stats (
	metric object(dynamic) as (
		"eventTimestamp" timestamp with time zone
	)
) clustered into 9 shards
with (number_of_replicas = 1);

```

```auto
create table test (
	metric object(dynamic) as (
		"eventTimestamp" timestamp with time zone
	)
) clustered into 9 shards
with (number_of_replicas = 1);

```

Every Object has more or less 20 field each.  
The tables has 3,5 Milions of row but we expect to reach something like 50 Milions.

What do you think?

Regards,  
S.

---

<div class="post-metadata">

**Author:** ![proddata](https://sea2.discourse-cdn.com/flex020/user_avatar/community.cratedb.com/proddata/32/1379_2.png) [@proddata](https://community.cratedb.com/u/proddata)\
**Post date:** [December 1, 2021, 10:16pm UTC](https://community.cratedb.com/t/high-heap-usage-after-a-while/897/4 "2021-12-01T22:16:38Z")

</div>

Could you run the following query:

```sql
SELECT count(*) FROM information_schema.columns
where table_name IN ('test','stats')

```

As mentioned above I would increase the heap to 50% of system memory (or increase the system memory).  
There is a certain minimum requirement to keep information about lucene indexes / tables in memory. This however largely various depending on the used schemas.