#prometheus (2019-03)

prometheus

Archive: https://archive.sweetops.com/prometheus/

2019-03-28

tamsky avatar
tamsky

random guess where CW-Exporter might be useful: ALB metrics

:--1:1

2019-03-25

Abel Luck avatar
Abel Luck

by native metrics, you mean the host metrics from node_exporter?

Abel Luck avatar
Abel Luck

what would those metrics be that you use CW exporter for?

2019-03-14

Abel Luck avatar
Abel Luck

anyone using prometheus to monitor an aws stack with the cloudwatch exporter? is it worth using?

Abel Luck avatar
Abel Luck

I’m also curious how costs compare to using cloudwatch alone vs the api request costs of the exporter

pecigonzalo avatar
pecigonzalo

We use CW-Exporter, but only for things we cant get native metrics for. I dont know if that helps you much

2019-03-13

pecigonzalo avatar
pecigonzalo


I think I’m having trouble understanding what “always gets the metrics” means for your situation.
You have only 1 PG, you do a failed deployment or it goes down for any reason, now any job pushing to PG cant push to it, because its down until you fix it/rollback/etc
Are you sure you’re using pushgateway to solve a problem for which it is designed & recommended ?
Yes im 100% sure, eg scheduled jobs like backups or other croned or event triggered things.
If the first one goes down, you shouldn’t care, assuming both are equally scraped by prometheus.
It does matter, because if you have 2 PG, unless you push the same data to both, by either a proxy/LB that replicates the data to both or by pushing to both from the clients, you have different data

pecigonzalo avatar
pecigonzalo

eg. a backup job runs, after it ran it pushes a metrics to PG through a LB.

  • PG1 gets the connection and stores the info
  • PG2 does not, as the LB only sent to PG1

PG1 goes down, for whatever reason, Prom starts reading from PG2, which doesnt contain the metric for the backup. now your metrics show odd stuff, and might trigger an alert about a missing backup in our case

tamsky avatar
tamsky

Let me respond to each of those separately:
You have only 1 PG,

Is something forcing you to create a single point of failure? By deploying 1 PG, you create a SPOF. Have you considered deploying N+1 ?
you do a failed deployment or it goes down for any reason, now any job pushing to PG cant push to it, because its down until you fix it/rollback/etc

If, as you say, you only have 1 PG and you accidentally deploy a “failed deployment” (read: “a blackhole for metrics”), yes, you’re going to lose data. Load balancers alone won’t help, but healthchecks do. If your infrastructure is always testing that your PG instances are able to accept POST requests, then you have some assurances that active PG are able to accept metrics.

pecigonzalo avatar
pecigonzalo


Is something forcing you to create a single point of failure?
Are you even reading my messages?! that is exactly what im doing, and asking how others are doing that….

do you even run PG? because tbh it seems you have no clue what you are talking about

pecigonzalo avatar
pecigonzalo

do you understand PG only keeps the data at the receiving end?

tamsky avatar
tamsky

yes

pecigonzalo avatar
pecigonzalo

and if you have 2, unless you send the data to the second nodea s well, you are not going to have redundant data

pecigonzalo avatar
pecigonzalo

To be honest, it seems like you are trolling at this point

tamsky avatar
tamsky

are you looking to use pushgateway as a continuous scoreboard? because that’s what it sounds like – but that’s not what it’s designed for.

pecigonzalo avatar
pecigonzalo

Again, do you even read?

tamsky avatar
tamsky

absolutely I read and have been trying to understand how you’re using it and why exactly you think it’s a problem to have different data on different PGs

pecigonzalo avatar
pecigonzalo

Why dont you explain how YOU run your N+1 reployment instead?

tamsky avatar
tamsky

sure

tamsky avatar
tamsky

I was in the middle of composing this when you replied, so I’ll send it now:
Yes im 100% sure, eg scheduled jobs like backups or other croned or event triggered things.

If your backups and crons run in the same place (host/instance) then I’d suggest using the Textfile collector [1] I’ve definitely used it to export the status of a backup script… the script would even update metrics during execution, eg: it updated a start_time metric before triggering the backup and additional detailed metrics at exit: {exitcode, completion_time, last_backup_size_bytes}

If using textfile metrics is not possible, that’s fine – just hoping to make you aware it is option, if you weren’t already.

[1] https://github.com/prometheus/node_exporter#textfile-collector

prometheus/node_exporter

Exporter for machine metrics. Contribute to prometheus/node_exporter development by creating an account on GitHub.

pecigonzalo avatar
pecigonzalo

Not an option, and the textfile option is just a hack tbh, you are basically building metric state in your node

tamsky avatar
tamsky

now you sound trollish

pecigonzalo avatar
pecigonzalo

but not an option, this is lambda

tamsky avatar
tamsky

ok great - thanks for sharing the environment in which you’re trying to operate

pecigonzalo avatar
pecigonzalo

im waiting for you N+1 deployment,instead of how to save metrics to disk

pecigonzalo avatar
pecigonzalo

pushing lambda/cron/job style metrics is EXACTLY what PG is there for

tamsky avatar
tamsky

sure.

tamsky avatar
tamsky

An ALB with N+1 PG’s in the targetgroup can receive metrics from your ephemeral jobs.

tamsky avatar
tamsky

Your prometheus server collects metrics from each PG in the TG individually – it does not scrape them via the ALB

tamsky avatar
tamsky

rewinding a bit from the scenario above – are you doing anything currently to manage the metrics lifecycle on your current PG’s ?

tamsky avatar
tamsky

do you delete metrics at all?

pecigonzalo avatar
pecigonzalo

On PG, no I do not delete them, but that has not been an issue nor the topic of my question

pecigonzalo avatar
pecigonzalo

now, in your scenario

pecigonzalo avatar
pecigonzalo

you have a lambda, that runs, pushes a eg, timestamp of lastsuccess run (like many metrics do) to PG, through LB. It only lands in 1 of the PG, lets say PG1 Now PG1 is down, your new metrics would go to PG2. Now you have an alert or want to check something like: ` ((time() - pushgateway_thismetric_last_success) / 3600) > 24` to check last daily backup did run

pecigonzalo avatar
pecigonzalo

wouldnt the 2PG metrics be 2 different metrics (due to the node they come from) and make it so you get a trigger

pecigonzalo avatar
pecigonzalo

because PG2 might not have data for pushgateway_thismetric_last_success yet

tamsky avatar
tamsky
prometheus/pushgateway

Push acceptor for ephemeral and batch jobs. Contribute to prometheus/pushgateway development by creating an account on GitHub.

tamsky avatar
tamsky

so the only situation where what you’ve described creates the situation you sound concerned about would be: your lambda pushes via the LB to pg1, after which pg1 dies without being scraped by your prometheus server. in all other situations where lambda pushes via the LB to pg1 and prometheus scrapes it, you’re safe.

pecigonzalo avatar
pecigonzalo

because of merges metrics+{labelsets} into the same metrics timeseries history?

tamsky avatar
tamsky

because your query for pushgateway_thismetric_last_success should return the latest metric, whether it arrived via pg1 or pg2

pecigonzalo avatar
pecigonzalo

Yeah, that makes sense, cool

pecigonzalo avatar
pecigonzalo

I was under the impression it would not do that due to instance being different basically

1
tamsky avatar
tamsky

you have to make sure the instance label is being set to ("") on scrape

pecigonzalo avatar
pecigonzalo

Yep, they act as basically the same metric

tamsky avatar
tamsky

that’s the behavior you want

pecigonzalo avatar
pecigonzalo

Indeed

tamsky avatar
tamsky

if you do have “can’t lose” metrics – you’ll want a more durable path to getting the metrics to prometheus

pecigonzalo avatar
pecigonzalo

Yeah, tbh in the rare case of the metric getting pushed to PG1 and prom not scraping in time

pecigonzalo avatar
pecigonzalo

i can tolerate the bad alert

pecigonzalo avatar
pecigonzalo

Or I guess there the persisting to disk or some other exporter would make more sense

tamsky avatar
tamsky

your questions here got me thinking of a setup : api gateway -> lambda sns writer -> sns readers -> dynamodb ) + ( /metrics scrape endpoint -> api gw -> lambda dynamo reader )

:--1:1
pecigonzalo avatar
pecigonzalo

that would be better than PG tbh

tamsky avatar
tamsky

right ? glad you agree

pecigonzalo avatar
pecigonzalo

you could even ALB -> lambda -> dynamo

tamsky avatar
tamsky

I searched around and didn’t find anything even discussing that setup

pecigonzalo avatar
pecigonzalo

without SNS and api gate

tamsky avatar
tamsky

I was thinking of a fanout for writes to go to multiple readers, in case you wanted cross-region HA

pecigonzalo avatar
pecigonzalo

I guess you could make it have 2 endpoints

pecigonzalo avatar
pecigonzalo

one for direct push, 1 for fanout or similar

pecigonzalo avatar
pecigonzalo

or make it extensible for fanout

tamsky avatar
tamsky

maybe there’s a market for an HA pushgateway

pecigonzalo avatar
pecigonzalo

as for many people ALB->lambda->dynamo might be enought

tamsky avatar
tamsky

I was also thinking that having TTLs on the dynamo entries would kind-of automate the metrics deletion aspects that you’ve waved your hands at

:--1:1
pecigonzalo avatar
pecigonzalo

if the entire ALB and lambda of region goes down, i have probably bigger issues

pecigonzalo avatar
pecigonzalo

Yeah, tbh we dont have that issue right now, as most of our metrics are not “stale”

tamsky avatar
tamsky

auto-metrics-deletion gives a lot of rope to folks to create high cardinality issues for themselves.

pecigonzalo avatar
pecigonzalo

but if you run most of your stuff in serverless, I guess you do have that problem

tamsky avatar
tamsky

running your stuff in serverless isn’t a fixed concern – it’s always how you’ve designed your labelsets. you have to be cognizant of their cardinality.

pecigonzalo avatar
pecigonzalo

Yeah, but pushed on button, as we only run satellite jobs and some simple (but important) metrics from them (backups, cleanups, checks, auto handle of failed events, etc) we dont have too many labels or cardinality

1
tamsky avatar
tamsky

you might benefit from having instance on a pushed metric in the pushgateway – so you can quickly find+debug the job that pushed it. but when you scrape it, that instance label is no longer useful, so it should be dropped/nulled.

pecigonzalo avatar
pecigonzalo

Yeah, that basically was my entire confusion with this PG stuff at the end

tamsky avatar
tamsky

yay!

tamsky avatar
tamsky

glad to help

pecigonzalo avatar
pecigonzalo

that is why I wanted to check how others were doing, because just discarding random stuff on the prometheus config seemed odd, but we needed a way to handle more than 1PG so we HA

tamsky avatar
tamsky

If you don’t yet, I’d suggest you run a blackbox prober scheduled lambda that pushes metrics on an interval, both individually to each PG (avoiding the ALB), as well as through the ALB. This gives you an alertable metric that your PG is not functioning.

pecigonzalo avatar
pecigonzalo

wouldnt this be covered by up()?

pecigonzalo avatar
pecigonzalo

but you are also correct, once Prom got the metric from one PG, and as long as your query “accepts” or “discards” lables that make the metrics from any of the PGs different, it should work

tamsky avatar
tamsky

yes, how you compose your alert rules is key

pecigonzalo avatar
pecigonzalo

ill do a local test with compose,thanks for the help

pecigonzalo avatar
pecigonzalo

glad we could understand each other at the end

tamsky avatar
tamsky

same here.

pecigonzalo avatar
pecigonzalo

I have one more @tamsky if you have a minute

pecigonzalo avatar
pecigonzalo

if you have PG1 last_success = 123 PG2 last_success = 124 (because the LB sends one to each) dont they get overwriten on the prom side after each scrape and make for odd metrics?

tamsky avatar
tamsky

timestamps should distinguish the two

tamsky avatar
tamsky

when you push, you push with a timestamp?

tamsky avatar
tamsky
prometheus/pushgateway

Push acceptor for ephemeral and batch jobs. Contribute to prometheus/pushgateway development by creating an account on GitHub.

tamsky avatar
tamsky

reading that, maybe I’m confusing it with the timestamps in the textfile exporter

tamsky avatar
tamsky
05:50:10 PM

rereads that link

tamsky avatar
tamsky

do you have a push_time_seconds for your last_success metrics group?

pecigonzalo avatar
pecigonzalo

push_time_seconds gets generated automatically

pecigonzalo avatar
pecigonzalo

afaik

pecigonzalo avatar
pecigonzalo

there is a timestime field tho, which in my test I did not push

pecigonzalo avatar
pecigonzalo
06:02:38 PM

push_time_second

pecigonzalo avatar
pecigonzalo
/tmp/tmp.Aj9RooPWfR on  master [!] on 🐳 v18.09.3 took 2s 
❯ echo "some_metric 3.14" | curl -u "admin:admin" --data-binary @- <http://localhost:9091/metrics/job/some_job>

/tmp/tmp.Aj9RooPWfR on  master [!] on 🐳 v18.09.3 
❯ echo "some_metric 3.15" | curl -u "admin:admin" --data-binary @- <http://localhost:9092/metrics/job/some_job>
pecigonzalo avatar
pecigonzalo
06:03:36 PM

the sample metric looks like

tamsky avatar
tamsky

I see what’s going on. thinking

pecigonzalo avatar
pecigonzalo

that is the problem i have with my backups metrics as well, that is why i was trying to basically “replicate” traffic to both PGs

pecigonzalo avatar
pecigonzalo

I might be missing something in the config tho idk

pecigonzalo avatar
pecigonzalo
  - job_name: "pushgateway"
    scrape_interval: 10s
    honor_labels: true
    static_configs:
      - targets: ["pushgateway1:9091", "pushgateway2:9091"]
pecigonzalo avatar
pecigonzalo

just in case

tamsky avatar
tamsky

what does a graph of push_time_seconds look like?

pecigonzalo avatar
pecigonzalo

the same

tamsky avatar
tamsky

should be two flat lines

pecigonzalo avatar
pecigonzalo

no, its just one

pecigonzalo avatar
pecigonzalo

because they dont have any labels that distinguishes them

tamsky avatar
tamsky

and you pushed at nearly the same time to both 9091 & 9092 ?

pecigonzalo avatar
pecigonzalo

some seconds after

pecigonzalo avatar
pecigonzalo

no, not at the same time

pecigonzalo avatar
pecigonzalo

If you have an LB one can come now, the other in 10 minutes and hit the other node

tamsky avatar
tamsky


that is the problem i have with my backups metrics as well, that is why i was trying to basically “replicate” traffic to both PGs

I understand that you want to try to replicate PUT/POST to the PG… but I’m now thinking I’m wrong about HA being a valid design pattern for this.

I was thinking that the PG timestamps operated similarly to the file exporter / text format

Exposition formats | Prometheus

An open-source monitoring system with a dimensional data model, flexible query language, efficient time series database and modern alerting approach.

tamsky avatar
tamsky

rewinding to the decision of “I need HA for PG” – what’s driving that decision?

pecigonzalo avatar
pecigonzalo

Dont you run PG in HA? and even you said
By deploying 1 PG, you create a SPOF. Have you considered deploying N+1 ?

pecigonzalo avatar
pecigonzalo

dont you have the same issue?

tamsky avatar
tamsky

I’ve run PG in multiple regions, but haven’t needed HA, single node ASG has been fine.

pecigonzalo avatar
pecigonzalo

So basically no

tamsky avatar
tamsky

it’s possible HA can still work, you’d add a non- job/instance label on scrape of each PG with a unique value and then can query for the most recent group of metrics across all scrapes

pecigonzalo avatar
pecigonzalo

Yeah

tamsky avatar
tamsky

and your metrics query looks like a join

pecigonzalo avatar
pecigonzalo

Now to Why, as explained in the question before

pecigonzalo avatar
pecigonzalo

when you asked that

pecigonzalo avatar
pecigonzalo

what if your deploy a bad version?

pecigonzalo avatar
pecigonzalo

etc

tamsky avatar
tamsky

yes.

pecigonzalo avatar
pecigonzalo

will help

tamsky avatar
tamsky

if you deploy a metrics blackhole, you’re going to lose pushes

pecigonzalo avatar
pecigonzalo

Im surprised im the only one having this “issue” with PG not supporting HA

tamsky avatar
tamsky

The thing you seem concerned with (“bad config/deploy”) is not unique to PG

tamsky avatar
tamsky

What if you deployed a bad version of prometheus? Or your lambda functions?

tamsky avatar
tamsky

But you might be able to get there with max_over_time(push_time_seconds{job="myjob"}[5m])

pecigonzalo avatar
pecigonzalo

that is why you run prom in HA

pecigonzalo avatar
pecigonzalo

with 2+ nodes and even federation

pecigonzalo avatar
pecigonzalo

and a DNS/Server/Client based balance for EG grafana

tamsky avatar
tamsky

yes, but nobody gets concerned if you lose one of 2 nodes, and that a new node doesn’t have exactly the same set of data that the older node does

pecigonzalo avatar
pecigonzalo

but quite similar

pecigonzalo avatar
pecigonzalo

its clearly not the same, prometheus has a pseudoHA config

tamsky avatar
tamsky

I don’t agree with the statement: prometheus has a pseudoHA config

pecigonzalo avatar
pecigonzalo

where you scrape same targets, and the only difference is due to timing, which you have to live with or not use prometheus

tamsky avatar
tamsky

that I agree with.

pecigonzalo avatar
pecigonzalo

that is their way of HA

tamsky avatar
tamsky

it’s redundancy

tamsky avatar
tamsky

when running multiple promservers nobody expects them to store identical timeseries history to disk

pecigonzalo avatar
pecigonzalo

And I never said that

tamsky avatar
tamsky

but that’s HA

pecigonzalo avatar
pecigonzalo

no

pecigonzalo avatar
pecigonzalo

Where does it say that is HA?

pecigonzalo avatar
pecigonzalo

that is maybe your personal take on HA

tamsky avatar
tamsky

that seemed to be the desired outcome for the PG – that an HA version of PG would replicate writes to all PGs

tamsky avatar
tamsky

that sort of replicated write does seem like a possible way to configure PG

pecigonzalo avatar
pecigonzalo

max_over_time(push_time_seconds[1m]) would work for evergrowing metrics, but not for a gauge

pecigonzalo avatar
pecigonzalo

if you push something like items_processed you can do that

pecigonzalo avatar
pecigonzalo

and you get a flapping metric again

tamsky avatar
tamsky

lets get an example going for max_over_time

tamsky avatar
tamsky

push_time_seconds{job=”myjob”,pg=”pg1”} 111 push_time_seconds{job=”myjob”,pg=”pg2”} 122

pecigonzalo avatar
pecigonzalo

where is pg= coming from?

pecigonzalo avatar
pecigonzalo

are you adding instance labels?

pecigonzalo avatar
pecigonzalo

because the client pushing, is not aware to which PG is pusshing

tamsky avatar
tamsky

I mentioned it earlier:
it’s possible HA can still work, you’d add a non- job/instance label on scrape of each PG with a unique value and then can query for the most recent group of metrics across all scrapes

pecigonzalo avatar
pecigonzalo

OK

pecigonzalo avatar
pecigonzalo

Yeah, Prom side

tamsky avatar
tamsky

yup

pecigonzalo avatar
pecigonzalo

yeah, doing a join does work, I meant direct max_overtime does not

pecigonzalo avatar
pecigonzalo

quite… hacky imho

tamsky avatar
tamsky

I’d think you’d use max_over_time to select the series and pg label that has the most recent timestamp (you’re running ntp right )

pecigonzalo avatar
pecigonzalo

yeah, that could work as a join, but its quite a hacky way as said before

tamsky avatar
tamsky

I agree hacky

tamsky avatar
tamsky

instead of engineering the PG to be HA have you considered making the PG clients more robust? what fallback mechanisms do you have if PG is down?

pecigonzalo avatar
pecigonzalo

Yes, I did, as explained when I started this, I could have 2 targets per job

tamsky avatar
tamsky

that’s just 2x fire-and-forget

tamsky avatar
tamsky

I’m thinking a separate backup path if the primary is down. Backup might not be HTTP.

pecigonzalo avatar
pecigonzalo

I think at that point i would rather add a proxy that replicates the data to multiple PGs

pecigonzalo avatar
pecigonzalo

as its a solution that applies to all my lambdas/jobs instead of an implementation per job

tamsky avatar
tamsky

how do you ameliorate the situation where the proxy misses a write to one of the PGs and the datasets diverge?

tamsky avatar
tamsky

you’ll wind up with a fluctuating graph, like you shared

tamsky avatar
tamsky

unless you run one prom server per PG ?

pecigonzalo avatar
pecigonzalo

but then its the same, 1 alrets one doesnt

pecigonzalo avatar
pecigonzalo

multiple prom will not fix it

pecigonzalo avatar
pecigonzalo

but yeah you are right, if the node returns, we might be back to square one

tamsky avatar
tamsky

I think “be careful” will still apply

pecigonzalo avatar
pecigonzalo

unless we have non persistent metrics

tamsky avatar
tamsky

I wasn’t thinking the PG node goes up/down/up

pecigonzalo avatar
pecigonzalo

which then the new pg does not have a metrics for the metrics that the other PG does

tamsky avatar
tamsky

I was simply thinking the proxy fails to communicate a PUT/POST for any number of network reasons – and it doesn’t retry, so that PUT/POST gets dropped on the floor.

pecigonzalo avatar
pecigonzalo

I would ask, how often is that bound to happen on an internal network?

tamsky avatar
tamsky

a non-zero number of times

tamsky avatar
tamsky

pecigonzalo avatar
pecigonzalo

I didnt ask how many times, I asked how often

tamsky avatar
tamsky

divide by choice of duration

pecigonzalo avatar
pecigonzalo

which is, if this affects only 1% of the cases, im not really that worried for this scneario

pecigonzalo avatar
pecigonzalo


how do you ameliorate the situation where the proxy misses a write to one of the PGs and the datasets diverge?
In a number of ways, for the number of problems

pecigonzalo avatar
pecigonzalo

-.-

tamsky avatar
tamsky

I’m thinking from all this that you have different classes of metrics that you’re pushing – some are frequent and business critical and are categorized “can’t lose”, and others are batch processes, like backups where the metric is useful but losing it isn’t intrinsically harmful.

is that correct?

pecigonzalo avatar
pecigonzalo

Well that is the thing, a metric for backups going missing is quite harmful depeding on the period

pecigonzalo avatar
pecigonzalo

as it could trigger a number of alerts

pecigonzalo avatar
pecigonzalo

OFC, if the soltuion to HA PG is more complex than just a couple of silences here and there

pecigonzalo avatar
pecigonzalo

ill just silence them when the event presents

pecigonzalo avatar
pecigonzalo

but Im suprised this is not a bigger problem for other usecases, eg. heavy serverless users

tamsky avatar
tamsky

triggering alerts definitely has a cost, but maybe there’s a different metric, (for example, via a blackbox prober) that reports the actual status of the latest backup

pecigonzalo avatar
pecigonzalo

you mean <https://github.com/prometheus/blackbox_exporter>?

tamsky avatar
tamsky

yes

pecigonzalo avatar
pecigonzalo

Yeah, i mean that is basically saying “dont use PG if you need PG HA” which is ok

pecigonzalo avatar
pecigonzalo

maybe PG is not the tool for this

pecigonzalo avatar
pecigonzalo

but im asking in a more generic way

tamsky avatar
tamsky

a blackbox example might be: a script that reads the state of all backups and reports the most recent timestamp

pecigonzalo avatar
pecigonzalo

how do you run PG in HA or some way of redundancy

pecigonzalo avatar
pecigonzalo

and there does not seem to be a way to do that

pecigonzalo avatar
pecigonzalo

Just to be clear, I know i can solve this in a number of different ways

pecigonzalo avatar
pecigonzalo

but in this case this is not my question, it is rather, how would you HA PG?

tamsky avatar
tamsky

then I agree, there is no out of box clustered PG that offers a singular HA datastore

pecigonzalo avatar
pecigonzalo

or replicated, even if delayed

pecigonzalo avatar
pecigonzalo

like gossip based for example

pecigonzalo avatar
pecigonzalo

Yeah, out of the box nothing, but maybe there are some clever ways to do this

tamsky avatar
tamsky

I’d use the existing singleton PG pattern, use ASG, and put it behind an ALB.

pecigonzalo avatar
pecigonzalo

EG: a replication proxy (altho your comment makes a fair point against it)

pecigonzalo avatar
pecigonzalo

Yeah, we do that already basically with a container

tamsky avatar
tamsky

I’d implement a backup channel in the clients to write metrics to SNS and have a persistent reader dequeue and PUT SNS data->PG

tamsky avatar
tamsky

clients would use the backup channel if the primary PG failed or timed out

pecigonzalo avatar
pecigonzalo

The thing is the problem is not the channel tbh, if that was the case, you could argue the same point about any HTTP communication

pecigonzalo avatar
pecigonzalo

but it could be a good solution to have some sort of queue

tamsky avatar
tamsky

it depends on the criticality of the data – if you have a requirement that you can never lose writes, you need more backup channels

pecigonzalo avatar
pecigonzalo

so client -> SNS/SQS/whatever -> deque -> PG

pecigonzalo avatar
pecigonzalo

so if PG goes down, it will just put the metrics once its back up

tamsky avatar
tamsky

I’d throw in a S3 writer if I thought SNS wasn’t reliable enough

pecigonzalo avatar
pecigonzalo

Yeah, i agree, but again, I guess you could say the same about any communication

tamsky avatar
tamsky

maybe an out-of-region backup PG ALB as an alternate

pecigonzalo avatar
pecigonzalo

and I dont see this being a pattern for any REST service for example

tamsky avatar
tamsky

you have this problem even in kafka writers

tamsky avatar
tamsky

kafka guarantees writes, but only after ACK. what do you do as a writer with critical data that doesn’t receive an ACK.

tamsky avatar
tamsky
  1. keep trying
  2. queue to a different system
  3. all of the above
tamsky avatar
tamsky
  1. fall over and ask for help
pecigonzalo avatar
pecigonzalo

yeah but you dont com over a different system

pecigonzalo avatar
pecigonzalo

normally, you retry or buffer and retry

pecigonzalo avatar
pecigonzalo

or fallover

tamsky avatar
tamsky

you don’t have as many options in lambda

pecigonzalo avatar
pecigonzalo

Exactly, but I think its not a great architecture patter to say

pecigonzalo avatar
pecigonzalo

“all lambdas HTTP to PG or send over SNS if that fails”

pecigonzalo avatar
pecigonzalo

because the same could be said about any HTTP connection in the same way

pecigonzalo avatar
pecigonzalo

waht about lambda calling another service?

pecigonzalo avatar
pecigonzalo

should it SNS as well?

pecigonzalo avatar
pecigonzalo

I would rather just make SNS or a queue the com channel

pecigonzalo avatar
pecigonzalo

and if that fails, then fallover and cry

tamsky avatar
tamsky

I can agree with that decision… just use SNS for all metrics writes

pecigonzalo avatar
pecigonzalo

(given you can async)

tamsky avatar
tamsky

HTTP over SNS

pecigonzalo avatar
pecigonzalo

Yeah, queuing the PG metrics might be a sane idea, that mitigate 90% of the issues, for not that much extra work

tamsky avatar
tamsky

actually I’m thinking SQS is the thing we’re talking about, not SNS

pecigonzalo avatar
pecigonzalo

Yeah, I said SNS as you said SNS before, but yeah im talking about a queue

tamsky avatar
tamsky

yeah - we’re talking same

pecigonzalo avatar
pecigonzalo


so client -
SNS/SQS/whatever -
deque -
PG
as said before

1
tamsky avatar
tamsky

the question isn’t how to handle REST failures in lambda, it’s how to deliver unique unreplicated data that will be otherwise lost unless written somewhere

pecigonzalo avatar
pecigonzalo

Yep, i think that is accurate

tamsky avatar
tamsky

my mistake using SNS

tamsky avatar
tamsky

and hopefully you can engineer it so that some amount of loss is acceptable

tamsky avatar
tamsky

as opposed to “no amount of loss is acceptable”

pecigonzalo avatar
pecigonzalo

Yeah, that i guess depends on the use case, as said for backups/simple stuff, it might be the silence-work<architectural-complexity

pecigonzalo avatar
pecigonzalo

but i was curious about how this is tackled across places, as i can think of many ways/things where this becomes more critical

tamsky avatar
tamsky

hear me out for another minute

tamsky avatar
tamsky

lets say your cron job that performs backups fails to report its status to PG

tamsky avatar
tamsky

and you have an alarm NoRecentBackupsHaveCompleted which fires after N+1 hours

tamsky avatar
tamsky

I think it’s also wise to have another job that reads the most recent backup and exports what it read.

tamsky avatar
tamsky

lets say that other “backup-reader” job has a metric alarm NoRecentValidBackup – which would be in state OK if the backup worked ok

tamsky avatar
tamsky

you can combine those two alerts to inhibit. you don’t need silences.

tamsky avatar
tamsky

lmk if you need more detail.

pecigonzalo avatar
pecigonzalo

I mean, yes this will fix it for backups, but while a second metric seems like a good idea for this case anywa, I think it’s a rather inelegant for the PG issue itself

tamsky avatar
tamsky


how this is tackled across places, as i can think of many ways/things where this becomes more critical

I haven’t worked with enough shops that were totally serverless to know the pitfalls. I do know some folks are very happy with couchbase for event metrics. I’ve also seen mondemand used as an event bus (in both multicast and anycast modes). At some point, durable logging becomes your primary concern, and metrics ingestion becomes secondary.

pecigonzalo avatar
pecigonzalo

Yeah, that is another option I considered, doing just logging for it

pecigonzalo avatar
pecigonzalo

And you could even do metric form that, like mtail

tamsky avatar
tamsky

That reminds me of one other excellent resource that I’ve used, and that’s nsq

tamsky avatar
tamsky

metrics/events can be written to nsq and tailed like you mentioned into different systems

pecigonzalo avatar
pecigonzalo

I’ll check it out

tamsky avatar
tamsky

nsq is pretty awesome

2019-03-12

pecigonzalo avatar
pecigonzalo

We do have a LB and DNS

pecigonzalo avatar
pecigonzalo

but the problem is that you need the data in both places

pecigonzalo avatar
pecigonzalo

is like having 2 postgres database, wihtout replication, and trying to read form the 2nd one if hte first one goes down, you dont have the same data

tamsky avatar
tamsky


to avoid downtime during deployments/etc so it always gets the metrics
EG: i roll a new update, config is bad, i broke the PG, and now Im losing all PG metrics coming in (backups, etc)
if this was HA, a rolling/blue-green would be really simple

I think I’m having trouble understanding what “always gets the metrics” means for your situation.
is like having 2 postgres database, wihtout replication, and trying to read form the 2nd one if hte first one goes down, you dont have the same data

Are you sure you’re using pushgateway to solve a problem for which it is designed & recommended ?

If not, please describe what you’re trying to solve for.

When to use the Pushgateway | Prometheus

An open-source monitoring system with a dimensional data model, flexible query language, efficient time series database and modern alerting approach.

tamsky avatar
tamsky


if hte first one goes down, you dont have the same data

If the first one goes down, you shouldn’t care, assuming both are equally scraped by prometheus. Prometheus reads the metrics from the first (down) and second (up) and merges metrics+{labelsets} into the same metrics timeseries history.

2019-03-11

tamsky avatar
tamsky

@pecigonzalo what are your goals with running pushgateway with an HA setup?

pecigonzalo avatar
pecigonzalo

to avoid downtime during deployments/etc so it always gets the metrics

pecigonzalo avatar
pecigonzalo

EG: i roll a new update, config is bad, i broke the PG, and now Im losing all PG metrics coming in (backups, etc)

pecigonzalo avatar
pecigonzalo

if this was HA, a rolling/blue-green would be really simple

tamsky avatar
tamsky

have you considered using either DNS + a load balancer – or service discovery in clients – which should help?

tamsky avatar
tamsky

I’ve never run PG in production, but if I did, I would avoid running it as a singleton. Is there anything non-trivial involved with running more than one PG (replicas with load-balancing, healthchecks)?

2019-03-07

pecigonzalo avatar
pecigonzalo

Anyone running Pushgateway in HA“HA”?

pecigonzalo avatar
pecigonzalo

We are evaluating multiple options, but could not find a lot of info about it, as I can see we:

  • Replicate traffic to multiple PG from another proxy
  • Send reports to multiple PGs

2019-03-04

Glenn J. Mason avatar
Glenn J. Mason

Hey folks, I’m deploying Prometheus (with helm and the prometheus-operator chart from stable) into several clusters, but wondering how best to set up the alerting rules? I’ve added a Slack receiver to Alertmanager, and I’m getting alerts there, but … any clues on the process I would use to ensure good alerting on my clusters?

Erik Osterman (Cloud Posse) avatar
Erik Osterman (Cloud Posse)

Hrmm @Glenn J. Mason can you add some more context around the process piece?

Erik Osterman (Cloud Posse) avatar
Erik Osterman (Cloud Posse)

Process for updating alerts?

Erik Osterman (Cloud Posse) avatar
Erik Osterman (Cloud Posse)

Where to define them?

Glenn J. Mason avatar
Glenn J. Mason

More like … how do I go about discovering what metrics mean “bad, or going bad”? Like … do I just monitor in e.g. grafana and/or make up “sensible” defaults that I think might work? The human process for tuning I mean.

Erik Osterman (Cloud Posse) avatar
Erik Osterman (Cloud Posse)

Have you started by looking over the many community provided dashboards?

Glenn J. Mason avatar
Glenn J. Mason

Yep, looking, but I was really wondering about the Prometheus alerting rules rather than Grafana dashboards specifically. I think the default rules are pretty good, but (e.g.) we’ve got the micrometer.io actuator getting JVM metrics in, and it’s hard to really be able to say “yep, that’s healthy” — except of course by monitoring and historical trends, etc. Which is certainly useful, and the “normal” way to do these things.

Glenn J. Mason avatar
Glenn J. Mason

I think it’s the only way to go, perhaps: monitor, decide on “normal” and take it from there.

:--1:1
Erik Osterman (Cloud Posse) avatar
Erik Osterman (Cloud Posse)

Yea, you’ll need to establish a baseline before you’ll really know what to alert on

    keyboard_arrow_up