TPU Site issues 2016/17

[W1zzard]

Galera MySQL database just keeps on crashing after a few minutes

 

[jboydgolfer]

 

[W1zzard]

Found an issue related to docker+overlayfs leading to container filesystem corruption. Fingers crossed that it works better now,

 

[FireFox]

Yup, too many things to fix. Any experts for centos+docker+pacemaker+glusterfs+nginx+galera+redis+memcached?

Nah, a Bag of spicy Mexican Chips and Titanfall2 online mode is my thing.

 

[dorsetknob]

I got that page to show up a couple of times.

I saw it once>>>>>> in all the time the site was down

And i thought Generously Wizz is trying to show off his later work

 

[qubit]

Found an issue related to docker+overlayfs leading to container filesystem corruption. Fingers crossed that it works better now,

Could you actually have an underlying hardware failure maybe? I've never seen TPU plagued by so many problems.

 

[W1zzard]

Could you actually have an underlying hardware failure maybe? I've never seen TPU plagued by so many problems.

On multiple machines at the same time?

 

[qubit]

On multiple machines at the same time?

No. Perhaps failover isn't working properly? Without knowing more about your setup I can only guess.

Think back: do you remember problems starting after any particular update or upgrade?

 

[Easy Rhino]

Galera is not enterprise ready. I don't care what anyone says. Also, what do you have running in docker? Don't tell me nginx...

 

[qubit]

Don't tell me nginx...

Is it a bit rubbish then? I remember another site running on it a few years ago tended to have problems.

 

[Easy Rhino]

Is it a bit rubbish then? I remember another site running on it a few years ago tended to have problems.

i simply do not see the point in running something like nginx in a container. it is so light weight that the only reason you should contain it is if you are concerned users/processes sharing the host could somehow compromise the instance. let me put it this way, containers are only useful if they are deployed into a cloud space web service. if you have a VM then you dont need to setup containers inside the VM. i mean, people do it because it is new and fun but it is the wrong tool for the job.

 

[W1zzard]

Galera is not enterprise ready. I don't care what anyone says. Also, what do you have running in docker? Don't tell me nginx...

Everything except for gluster is running in containers.

Any alternatives to Galera? (that's not classic MySQL master-slave replication)

 

[BiggieShady]

Everything except for gluster is running in containers.

Any alternatives to Galera? (that's not classic MySQL master-slave replication)

I don't know how you run your nodes in a cluster, but I thought this might be relevant (from stackoverflow):

Galera replication is synchronous and when a write query gets back to you, the data is on every node in the cluster. If for any reason a node cannot apply the transaction, it will kill itself to maintain the integrity of data throughout the cluster.

it will kill itself means it will shutdown itself?

Yes, it will gracefully shut itself down.

I'm not an expert in this particular field, but maybe something asynchronous like https://github.com/vmware/tungsten-replicator

 

[W1zzard]

If for any reason a node cannot apply the transaction, it will kill itself to maintain the integrity of data throughout the cluster.

That's to be expected and at some level what happened. Better than the node keeps accepting writes and ends up with corrupted data.

 

[BiggieShady]

That's to be expected and at some level what happened. Better than the node keeps accepting writes and ends up with corrupted data.

Ah then, as far as I know anything usable for synchronous mode (no data loss) is Galera and NDB

 

[HTC]

Got this yesterday, when i lost access to TPU (see attach).

For a while, could not access TPU and got that every time, but it didn't last long: took a pic in case it may help, somehow, to prevent it happening again.

 

[W1zzard]

Got this yesterday, when i lost access to TPU (see attach).

Yeah the database was crashed

 

[Aenra]

My only site-related issue is that everyone has been like, super helpful to me and there's no way to give back

While i can understand how others may feel differently, i would suggest the inclusion of a donating function.
(unless there already is one? Got my blonde moments just like everyone else)

 

[FireFox]

 

[W1zzard]

I'm doing rolling server reboots right now, some slight hiccups (a few seconds) are to be expected

Edit: all finished

 

[jboydgolfer]

My only site-related issue is that everyone has been like, super helpful to me and there's no way to give back

While i can understand how others may feel differently, i would suggest the inclusion of a donating function.
(unless there already is one? Got my blonde moments just like everyone else)

there are giveaways on TPU ,hosted by TPU members for members in need, keep Your eyes peeled, and Im certain if your still motivated to "give back" you'll have the opportunity.
Keep your eyes peeled.

 

[Easy Rhino]

Any alternatives to Galera? (that's not classic MySQL master-slave replication)

Unfortunately no unless you want to use Percona's binaries. Are you in AWS using EC2 or using a hosting provider like RackSpace with VMs?

 

[W1zzard]

Unfortunately no unless you want to use Percona's binaries. Are you in AWS using EC2 or using a hosting provider like RackSpace with VMs?

Using bare-metal

 

[HTC]

Dunno if it's just my system but pic loading from TPU is very sloooooooooow.

For example: the pic in post #419 (from Knoxx29's post) takes around 5 seconds to finish displaying. For a small pic (in size), i find that quite unusual.

 

[W1zzard]

Loads nearly instantyl for me