You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
type FieldCodec struct {
fieldsByID map[uint8]*Field
fieldsByName map[string]*Field
}
fieldsByID is a map struct, the key type is uint8. That means only 255 fields are stored in fieldsByID.
I remember that FieldCodec is used in 0.9.x version based BoltDB storage engine with 255 fields limitation. Now, only some influx_inspect code use FieldCodec info and MeasurementFields. So, could we delete the FieldCodec ?
The is a another bug about FieldCodec as following:
Steps to reproduce: Writing new points(in only one series) with new fields continuously, the wait time is longer and longer and the CPU usage is quickly increase.
Now I have two method fix the bug, deleting the FieldCodec or fix the trouble code.If you can give me some suggestion that whitch method is better, I would like to focus on the issue and open a Pull Request.
The text was updated successfully, but these errors were encountered:
Updated `influx_inspect` to use the `FieldDimensions` method instead
(more reliable anyway). The `influx_tsm` program used its own vendored
copy of `FieldCodec` so it is not affected by this change. `FieldCodec`
was only used for the `b1` and `bz1` engines which were removed in 0.12,
but the code that created the field codec was never removed. This
limited the maximum number of fields to 255 even though that restriction
was removed with the `tsm1` engine.
Fixes#6869.
The
FieldCodec
type declaration as followinghttps://github.com/influxdata/influxdb/blob/master/tsdb/shard.go#L799-L802
fieldsByID
is a map struct, the key type isuint8
. That means only 255 fields are stored infieldsByID
.I remember that
FieldCodec
is used in0.9.x
version based BoltDB storage engine with255 fields limitation
. Now, only some influx_inspect code use FieldCodec info and MeasurementFields. So, could we delete theFieldCodec
?The is a another bug about FieldCodec as following:
Bug report
System info: [InfluxDB : 551cb61 ]
Steps to reproduce: Writing new points(in only one series) with new fields continuously, the wait time is longer and longer and the CPU usage is quickly increase.
The trouble code cause the bug.
Now I have two method fix the bug, deleting the
FieldCodec
or fix the trouble code.If you can give me some suggestion that whitch method is better, I would like to focus on the issue and open a Pull Request.The text was updated successfully, but these errors were encountered: