You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@trafficserver.apache.org by "ksqrtr (via GitHub)" <gi...@apache.org> on 2025/03/04 03:40:18 UTC
[I] Unbalanced NVME disks' usage [trafficserver]
ksqrtr opened a new issue, #12069:
URL: https://github.com/apache/trafficserver/issues/12069
Hello!
We have 4 NVME disks and noticed recently that their usage is some kind of unbalanced:
```sh
# for nvme in 0 1 2 3; do echo "nvme${nvme}:"; smartctl -x /dev/nvme"${nvme}"n1 | grep 'Data Units' | sed 's/^/ /' ; done
nvme0:
Data Units Read: 127 608 511 556 [65,3 PB]
Data Units Written: 293 649 334 [150 TB]
nvme1:
Data Units Read: 129 983 891 432 [66,5 PB]
Data Units Written: 384 164 362 [196 TB]
nvme2:
Data Units Read: 562 268 [287 GB]
Data Units Written: 8 382 370 [4,29 TB]
nvme3:
Data Units Read: 551 991 [282 GB]
Data Units Written: 8 387 551 [4,29 TB]
```
What might be the cause of such usage unbalance and is there any options in ATS we can tune?
Our `storage.config` is pretty straightforward:
```
# cat storage.config
/dev/crypt-nvme-luks-S2UDNX0J602140
/dev/crypt-nvme-luks-S2UDNX0J402480
/dev/crypt-nvme-luks-S2UDNX0J602013
/dev/crypt-nvme-luks-S2UDNX0J602462
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscribe@trafficserver.apache.org.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] Unbalanced RAW NVME disks usage [trafficserver]
Posted by "github-actions[bot] (via GitHub)" <gi...@apache.org>.
github-actions[bot] commented on issue #12069:
URL: https://github.com/apache/trafficserver/issues/12069#issuecomment-4035721850
This issue has been automatically marked as stale because it has not had recent activity. Marking it stale to flag it for further consideration by the community.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscribe@trafficserver.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] Unbalanced RAW NVME disks usage [trafficserver]
Posted by "cmcfarlen (via GitHub)" <gi...@apache.org>.
cmcfarlen commented on issue #12069:
URL: https://github.com/apache/trafficserver/issues/12069#issuecomment-2711987566
There could be a few reasons for this. Could you share your cache config and volume config along with some sample urls? If the same urls are updated frequently then they would all write to the same disk.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscribe@trafficserver.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] Unbalanced RAW NVME disks usage [trafficserver]
Posted by "bryancall (via GitHub)" <gi...@apache.org>.
bryancall closed issue #12069: Unbalanced RAW NVME disks usage
URL: https://github.com/apache/trafficserver/issues/12069
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscribe@trafficserver.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] Unbalanced RAW NVME disks usage [trafficserver]
Posted by "bryancall (via GitHub)" <gi...@apache.org>.
bryancall commented on issue #12069:
URL: https://github.com/apache/trafficserver/issues/12069#issuecomment-4828270912
I'm closing this as a stale support request. Traffic Server distributes cached objects across cache stripes (and therefore across disks) by hashing the object's content key, so write and read volume is only as balanced as the distribution of the URLs being cached. If a relatively small set of hot or frequently-updated URLs dominates your traffic, those objects hash to the same stripe and concentrate writes on one or two disks, which matches the large skew you saw between nvme0/nvme1 and nvme2/nvme3. An uneven storage.config or volume.config mapping can produce the same effect. Chris asked for your cache config, volume config, and some sample URLs to dig in further, but we did not hear back, so there is not enough information to investigate this as a bug. If you are still seeing this on a current release, please open a new issue with your storage.config, volume.config, cache.config, and a representative set of request URLs and I would be happy to take another look.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscribe@trafficserver.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org