Multiple influxdb databases

[eschek87]

Is it possible to specify/define more than one influxdb database? At the moment i have created one database and the metrics from vnx and vplex are in this database.

And i have seen that the metrics did not have a "tag" name in influxdb:

show series key

2-1-A,class=vplex,name=vplex-test 2-1-B,class=vplex,name=vplex-test be-prt.read,class=vplex,name=vplex-test,parentName=2-1-A be-prt.read,class=vplex,name=vplex-test,parentName=2-1-B be-prt.write,class=vplex,name=vplex-test,parentName=2-1-A be-prt.write,class=vplex,name=vplex-test,parentName=2-1-B

[vzaigrin]

Unfortunately, all storage systems output to one database at InfluxDB. But, output from VPLEX and VNX wiil have different class tag.

[eschek87]

Ok thanks. But the biggest problem is that all values got saved as "string":

SHOW FIELD KEYS name: 2-1-A fieldKey fieldType

---

director.be-ops string director.be-ops-read string director.be-ops-write string director.be-read string director.be-write string director.busy string director.fe-ops string director.fe-ops-act string director.fe-ops-q string director.fe-ops-read string director.fe-ops-write string director.fe-read string director.fe-write string

[vzaigrin]

Yes, they are string. What type they should have?

[eschek87]

I think they should be "int" or? Cause if i try to create my query in influx i got the following:

SELECT mean(director.busy) FROM "2-1-A" ERR: unsupported mean iterator type: *influxql.stringInterruptIterator

Without "mean" i got all values from all the time.

[vzaigrin]

'Int' is not a good idea. A lot of metrics are float. I will change output to InfluxDB from String to Float.

[eschek87]

The same affects the vnx data in influxdb

[eschek87]

Is it possible to include: SP Read Cache State Enabled SP Write Cache State Enabled So with influxdb as datasource i could create a singlestat panel with the text

[vzaigrin]

Yes, it is possible just by changing config file.

For VNX1: In the vnxblock extractor common parameters add to the method cache this strings into paramlist: <param>SP Read Cache State</param> <param>SP Write Cache State</param>

For VNX2: In the vnxblock extractor common parameters add to the method cache2 this strings into paramlist: <param>SPA Read Cache State</param> <param>SPA Write Cache State</param> <param>SPB Read Cache State</param> <param>SPB Write Cache State</param> And change line cmd from "cache -sp -info -perfData" to "cache -sp -info"

But in this case it will be hard to change output type from 'String' to 'Float'.

[eschek87]

ok. i wanted to switch to influxdb cause it can store "strings" and graphite cannot. I tried it but i got:

in /var/log/carbon/listener.log ) received from client 127.0.0.1:49612, ignoring.block.cache.spa.ReadCacheState 1486647160 ) received from client 127.0.0.1:49612, ignoring.block.cache.spa.WriteCacheState 1486647160 ) received from client 127.0.0.1:49614, ignoring.block.cache.spa.ReadCacheState 1486647160 ) received from client 127.0.0.1:49614, ignoring.block.cache.spa.WriteCacheState 1486647160 ) received from client 127.0.0.1:49612, ignoring.block.cache.spb.ReadCacheState 1486647160 ) received from client 127.0.0.1:49612, ignoring.block.cache.spb.WriteCacheState 1486647160 ) received from client 127.0.0.1:49614, ignoring.block.cache.spb.ReadCacheState 1486647160 ) received from client 127.0.0.1:49614, ignoring.block.cache.spb.WriteCacheState 1486647160

[vzaigrin]

No, this is other problem. When you will run naviseccli command you will see that outputs are very different even for one command. Sometime it separate value by "=", sometime by ":" For getting values from naviseccli output I match numbers and discard all other. I forgot about this when I advised how to get status of the cache. I will change Collector to get not only numbers as output.

[vzaigrin]

I have updated collector. Now it outputs strings from VNXBlock. Numbers to InfluxDB outputs as Double, but you need clean old values from database as it remembers type of previous values.

[eschek87]

Thank you for the modification! I tried to find out what causes the error. I modified a file to get more information, but this did not helped me:

]/02/2017 20:56:24 :: invalid line - [ vnx.prz-vnx.block.cache.spa.ReadCacheState Enabled 1486673773 ) received from client 127.0.0.1:54744, ignoring.block.cache.spa.WriteCacheState Enabled 1486673773 ]/02/2017 20:56:24 :: invalid line - [ vnx.prz-vnx.block.cache.spa.WriteCacheState Enabled 1486673773 09/02/2017 20:56:25 :: MetricLineReceiver connection with 127.0.0.1:54740 closed cleanly ) received from client 127.0.0.1:54744, ignoring.block.cache.spb.ReadCacheState Enabled 1486673773 ]/02/2017 20:56:27 :: invalid line - [ vnx.prz-vnx.block.cache.spb.ReadCacheState Enabled 1486673773 ) received from client 127.0.0.1:54744, ignoring.block.cache.spb.WriteCacheState Enabled 1486673773 ]/02/2017 20:56:27 :: invalid line - [ vnx.prz-vnx.block.cache.spb.WriteCacheState Enabled 1486673773

Have you any idea?

[vzaigrin]

It's looks like Carbon doesn't accept Strings as values. In the InfluxDB I saw Strings for this values.

[eschek87]

Yes in influxdb it works. Is it possible to send the data to influxdb in another order/sequence? At the moment its in the following order: director.per-cpu-busy,class=vplex,name=vplex-test,parentName=1-1-A

So i need to do the query in the following way

But i wanted to do "from director-name" and as field key the per cpu busy.

[vzaigrin]

At the present time parentName is a 'director name'. I did this for unification. Do you need simple 'per-cpu-busy' (without 'deirector.') as a metric name?

[eschek87]

No. With director name is ok. But the metrics for the directors should be stored under the director name as a field key. I think using graphite is easier. Thanks for your help