[storage_usage]One node occupied much more than the others

[weikang]

Please refer below, the first node is used as much as twice of the other nodes. Why isn't it evenly distributed? Thanks.

select node_name, used_bytes, usage_percent from storage_usage where filesystem='vertica' order by usage_percent desc;

node_name used_bytes usage_percent
v_stats_smry_node0001 1082878177280 35.33
v_stats_smry_node0010 510226530304 17.24
v_stats_smry_node0011 510059966464 17.24
v_stats_smry_node0005 508578189312 17.19
v_stats_smry_node0020 508723236864 17.19
v_stats_smry_node0016 507932114944 17.16
v_stats_smry_node0003 507402616832 17.15
v_stats_smry_node0017 507453112320 17.15
v_stats_smry_node0002 507187027968 17.14
v_stats_smry_node0019 507365732352 17.14
v_stats_smry_node0006 506279739392 17.11
v_stats_smry_node0014 505718636544 17.09
v_stats_smry_node0007 505431814144 17.08
v_stats_smry_node0009 505353199616 17.08
v_stats_smry_node0008 504889966592 17.06
v_stats_smry_node0004 497900834816 16.82
v_stats_smry_node0012 497286971392 16.8
v_stats_smry_node0018 489397899264 16.54
v_stats_smry_node0013 487197736960 16.46
v_stats_smry_node0015 486295859200 16.43

[NorbertKrupa]

What version of Vertica are you running? My initial reaction would be data skew or a projection incorrectly segmented.

[weikang]

Thanks for replying.

We are running 7.0.1. We are using segments: SEGMENTED BY MODULARHASH (C1,C2,C3) ALL NODES

What version of Vertica are you running? My initial reaction would be data skew or a projection incorrectly segmented.

[NorbertKrupa]

Can you run the query below and post the output or confirm that the segments are not skewed?

Code: Select all

SELECT segment_layout FROM v_internal.vs_elastic_cluster;

[scutter]

storage_usage is pure filesystem usage. Check projection_usage to see if the skew is within the projections or just on disk. If it’s on disk it maybe log files or CopyErrorLogs.