Re: [SLUG] assessing vps performance issues
Jeremy, Yes, more than enough . But, If you wanted to ask 'so who is the real idiot?', well, the answer is clear it's me. In my defense, I've dealt with these people for well over 10 years, though not for hosting, and, the vps performed adequately for 1.5 years, with only 1 glitch I think.( though the Win MS SQL server was a joke from day 1, I guess I should've learnt from that debacle) Two weeks into this incident, today got this: - Please find attached our IR for the most recent issue we experienced with our infrastructure. The issue experienced was directly related to maintenance work being conducted on the underlying storage. As a result maintenance tasks have been reviewed to ensure the do not create excessive load that will be visible to the customer environment. In addition it has also become clear that customers may also be able to generate high IO load in a similar manner to this maintenance with the potential to affect other clients systems. As a result we have now implemented hard IO limits on all VPS in the cluster (200IOPS) If you would like any further information please respond to this email. VPS Service Desk - Thanks again to everyone Jeremy Visser jer...@visser.name wrote: On 01/11/13 21:56, li...@sbt.net.au wrote: It only took them one week of on/off outages to arrive at that, I guess they must've been very thorough in reviewing it. and, about an hour AFTER I got that email, senior idiot phoned me to tell me I've loaded vps in excess of it,s capacity. So given all that, I think you have collected enough data to warrant changing providers. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html -- Sent from my Android phone with K-9 Mail. Please excuse my brevity. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
David, You are either very good at guessing, or perhaps you are not guessing... David Bomba turbo...@gmail.com wrote: Let me guess, Netregistry and their awesome support…. right? On 01/11/2013, at 9:56 PM, li...@sbt.net.au wrote: Thanks for the link, thanks for all the tips and advice, it's what kept me going through: 'we can ping it, why do you think there is a problem?' 'reboot it' 'you need more memory' 'don't you have a cron job every 5 minutes? that's the cause' I now received this -- Our System Administrators have reviewed Sar logs on his system and can see that increase IOwait can be seen at times that are in line with storage upgrade operation we have been conducting over the last few days. It was expected that storage vmotion of VM's during these upgrades would increase IO on the storage cluster however it was unanticipated that there would be any noticeable affect to our clients. We apologies for the interruption will process an SLA rebate for you as per our terms and would like to advise that this storage upgrade activity was completed Monday night. If you notice any further issues, please create a new ticket with details including output from Sar Logs indicating the time and date as well as the load experienced. -- It only took them one week of on/off outages to arrive at that, I guess they must've been very thorough in reviewing it. and, about an hour AFTER I got that email, senior idiot phoned me to tell me I've loaded vps in excess of it,s capacity. Thanks for both technical and mental support, guys. {And, they still havent fixed it, last night vps again overloaded} Amos Shapira amos.shap...@gmail.com wrote: In addition to Michael's good advise, see http://blog.scoutapp.com/articles/2013/07/25/understanding-cpu-steal-time-when-should-you-be-worriedabout steal time. -- Sent from my Android phone with K-9 Mail. Please excuse my brevity. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html -- Sent from my Android phone with K-9 Mail. Please excuse my brevity. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
On 5 Nov 2013, at 3:58 pm, Voytek Eymont li...@sbt.net.au wrote: In addition it has also become clear that customers may also be able to generate high IO load in a similar manner to this maintenance with the potential to affect other clients systems. As a result we have now implemented hard IO limits on all VPS in the cluster (200IOPS) Are you sure 200 IOPS isn’t a typo? If not, you are being ripped off. I use Linode, for example, and I get *warning* alerts (not a hard limit) by default at 1000 IOPS, and even that warning threshold is configurable. Linode has been benchmarked at much better than 200 IOPS: http://serverbear.com/10-linode-1gb-linode Even though on that ServerBear site, Linode doesn’t score particularly high in terms of I/O, I can tell you now, the crap that you’ve gone through generally Just Doesn’t Happen™. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
200 IOPS is criminal, another well known provider is 'promoting' their SSD backed servers with a default IOP of 100 why bother having SSD's when it will perform worse than a single 10k SAS drive On 5 November 2013 17:13, Jeremy Visser jer...@visser.name wrote: On 5 Nov 2013, at 3:58 pm, Voytek Eymont li...@sbt.net.au wrote: In addition it has also become clear that customers may also be able to generate high IO load in a similar manner to this maintenance with the potential to affect other clients systems. As a result we have now implemented hard IO limits on all VPS in the cluster (200IOPS) Are you sure 200 IOPS isn’t a typo? If not, you are being ripped off. I use Linode, for example, and I get *warning* alerts (not a hard limit) by default at 1000 IOPS, and even that warning threshold is configurable. Linode has been benchmarked at much better than 200 IOPS: http://serverbear.com/10-linode-1gb-linode Even though on that ServerBear site, Linode doesn’t score particularly high in terms of I/O, I can tell you now, the crap that you’ve gone through generally Just Doesn’t Happen™. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
Thanks for the link, thanks for all the tips and advice, it's what kept me going through: 'we can ping it, why do you think there is a problem?' 'reboot it' 'you need more memory' 'don't you have a cron job every 5 minutes? that's the cause' I now received this -- Our System Administrators have reviewed Sar logs on his system and can see that increase IOwait can be seen at times that are in line with storage upgrade operation we have been conducting over the last few days. It was expected that storage vmotion of VM's during these upgrades would increase IO on the storage cluster however it was unanticipated that there would be any noticeable affect to our clients. We apologies for the interruption will process an SLA rebate for you as per our terms and would like to advise that this storage upgrade activity was completed Monday night. If you notice any further issues, please create a new ticket with details including output from Sar Logs indicating the time and date as well as the load experienced. -- It only took them one week of on/off outages to arrive at that, I guess they must've been very thorough in reviewing it. and, about an hour AFTER I got that email, senior idiot phoned me to tell me I've loaded vps in excess of it,s capacity. Thanks for both technical and mental support, guys. {And, they still havent fixed it, last night vps again overloaded} Amos Shapira amos.shap...@gmail.com wrote: In addition to Michael's good advise, see http://blog.scoutapp.com/articles/2013/07/25/understanding-cpu-steal-time-when-should-you-be-worriedabout steal time. -- Sent from my Android phone with K-9 Mail. Please excuse my brevity. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
On 01/11/13 21:56, li...@sbt.net.au wrote: It only took them one week of on/off outages to arrive at that, I guess they must've been very thorough in reviewing it. and, about an hour AFTER I got that email, senior idiot phoned me to tell me I've loaded vps in excess of it,s capacity. So given all that, I think you have collected enough data to warrant changing providers. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
Let me guess, Netregistry and their awesome support…. right? On 01/11/2013, at 9:56 PM, li...@sbt.net.au wrote: Thanks for the link, thanks for all the tips and advice, it's what kept me going through: 'we can ping it, why do you think there is a problem?' 'reboot it' 'you need more memory' 'don't you have a cron job every 5 minutes? that's the cause' I now received this -- Our System Administrators have reviewed Sar logs on his system and can see that increase IOwait can be seen at times that are in line with storage upgrade operation we have been conducting over the last few days. It was expected that storage vmotion of VM's during these upgrades would increase IO on the storage cluster however it was unanticipated that there would be any noticeable affect to our clients. We apologies for the interruption will process an SLA rebate for you as per our terms and would like to advise that this storage upgrade activity was completed Monday night. If you notice any further issues, please create a new ticket with details including output from Sar Logs indicating the time and date as well as the load experienced. -- It only took them one week of on/off outages to arrive at that, I guess they must've been very thorough in reviewing it. and, about an hour AFTER I got that email, senior idiot phoned me to tell me I've loaded vps in excess of it,s capacity. Thanks for both technical and mental support, guys. {And, they still havent fixed it, last night vps again overloaded} Amos Shapira amos.shap...@gmail.com wrote: In addition to Michael's good advise, see http://blog.scoutapp.com/articles/2013/07/25/understanding-cpu-steal-time-when-should-you-be-worriedabout steal time. -- Sent from my Android phone with K-9 Mail. Please excuse my brevity. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
Hi, I would like to recommend Micron21 in Melbourne as a great hosting provider. I am one of their clients. Ben On 01/11/13 21:56, li...@sbt.net.au wrote: Content preview: Thanks for the link, thanks for all the tips and advice, it's what kept me going through: 'we can ping it, why do you think there is a problem?' 'reboot it' 'you need more memory' 'don't you have a cron job every 5 minutes? that's the cause' [...] Content analysis details: (-1.1 points, 5.0 required) pts rule name description -- -- 0.8 SPF_NEUTRALSPF: sender does not match SPF record (neutral) -1.9 BAYES_00 BODY: Bayes spam probability is 0 to 1% [score: 0.] X-Spam-Flag: NO Thanks for the link, thanks for all the tips and advice, it's what kept me going through: 'we can ping it, why do you think there is a problem?' 'reboot it' 'you need more memory' 'don't you have a cron job every 5 minutes? that's the cause' I now received this -- Our System Administrators have reviewed Sar logs on his system and can see that increase IOwait can be seen at times that are in line with storage upgrade operation we have been conducting over the last few days. It was expected that storage vmotion of VM's during these upgrades would increase IO on the storage cluster however it was unanticipated that there would be any noticeable affect to our clients. We apologies for the interruption will process an SLA rebate for you as per our terms and would like to advise that this storage upgrade activity was completed Monday night. If you notice any further issues, please create a new ticket with details including output from Sar Logs indicating the time and date as well as the load experienced. -- It only took them one week of on/off outages to arrive at that, I guess they must've been very thorough in reviewing it. and, about an hour AFTER I got that email, senior idiot phoned me to tell me I've loaded vps in excess of it,s capacity. Thanks for both technical and mental support, guys. {And, they still havent fixed it, last night vps again overloaded} Amos Shapira amos.shap...@gmail.com wrote: In addition to Michael's good advise, see http://blog.scoutapp.com/articles/2013/07/25/understanding-cpu-steal-time-when-should-you-be-worriedabout steal time. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
In addition to Michael's good advise, see http://blog.scoutapp.com/articles/2013/07/25/understanding-cpu-steal-time-when-should-you-be-worriedabout steal time. On 30 October 2013 12:28, Michael Chesterton che...@chesterton.id.auwrote: show the support people your evidence of high IO wait. if they can't fix it, change providers. Also if they have a public forum, ask for help on their public forum with the output showing high IO. If they're an au company, ask for help on whirlpool in the appropriate forum, they probably watch that for mentions of their name. On Wed, Oct 30, 2013 at 12:17 PM, li...@sbt.net.au wrote: On Fri, October 25, 2013 1:36 pm, Michael Chesterton wrote: I know it's resolved, but look at the 6th column in the original post. 90+% iowait, 90% of the time, the cpu is waiting for IO. now what you should do is run the same command when everything is working normally so you'll know what the output looks like. Then when you have another problem, you can compare with the normal output. -- the resolved problem keeps coming back, under normal situation, I'm ruining load average of .2 then, for maybe 1 to 2 hours high iowait, load average peaks at 130+ aprt from sar and bonnie (haven't installed it yet, wasn't in my repo), what other indicators can I use ? I'm thinking of feeding it into cacti. meanwhile, data centre response has been: 'we can ping it, there is nothing wrong' and 'reboot the server' -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html -- [image: View my profile on LinkedIn] http://www.linkedin.com/in/gliderflyer -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
On Fri, October 25, 2013 1:36 pm, Michael Chesterton wrote: I know it's resolved, but look at the 6th column in the original post. 90+% iowait, 90% of the time, the cpu is waiting for IO. now what you should do is run the same command when everything is working normally so you'll know what the output looks like. Then when you have another problem, you can compare with the normal output. -- the resolved problem keeps coming back, under normal situation, I'm ruining load average of .2 then, for maybe 1 to 2 hours high iowait, load average peaks at 130+ aprt from sar and bonnie (haven't installed it yet, wasn't in my repo), what other indicators can I use ? I'm thinking of feeding it into cacti. meanwhile, data centre response has been: 'we can ping it, there is nothing wrong' and 'reboot the server' -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
show the support people your evidence of high IO wait. if they can't fix it, change providers. Also if they have a public forum, ask for help on their public forum with the output showing high IO. If they're an au company, ask for help on whirlpool in the appropriate forum, they probably watch that for mentions of their name. On Wed, Oct 30, 2013 at 12:17 PM, li...@sbt.net.au wrote: On Fri, October 25, 2013 1:36 pm, Michael Chesterton wrote: I know it's resolved, but look at the 6th column in the original post. 90+% iowait, 90% of the time, the cpu is waiting for IO. now what you should do is run the same command when everything is working normally so you'll know what the output looks like. Then when you have another problem, you can compare with the normal output. -- the resolved problem keeps coming back, under normal situation, I'm ruining load average of .2 then, for maybe 1 to 2 hours high iowait, load average peaks at 130+ aprt from sar and bonnie (haven't installed it yet, wasn't in my repo), what other indicators can I use ? I'm thinking of feeding it into cacti. meanwhile, data centre response has been: 'we can ping it, there is nothing wrong' and 'reboot the server' -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
On Fri, October 25, 2013 3:25 pm, li...@sbt.net.au wrote: On Fri, October 25, 2013 1:36 pm, Michael Chesterton wrote: I know it's resolved, but look at the 6th column in the original post. 90+% iowait, 90% of the time, the cpu is waiting for IO. Michael, thanks again this morning I got a 'brief outage' with monitoring down/up advices like below, looking at cacti 'load' chart, drop in data output around 5am, followed by big spike around 6am when graph re-starts (size of spike might be the dropped data from 5am till graph resumed, perhaps) and, iowait again: --- # sar -s 01:00:00 -e 07:30:00 Linux 2.6.32-358.23.2.el6.x86_64 10/26/2013 _x86_64_(2 CPU) 01:00:01 AM CPU %user %nice %system %iowait%steal %idle 01:10:02 AM all 4.04 0.00 0.92 28.83 0.00 66.20 01:20:02 AM all 6.43 0.01 1.29 33.42 0.00 58.85 01:30:01 AM all 5.88 0.01 1.24 26.82 0.00 66.05 01:40:01 AM all 4.13 0.00 0.91 2.08 0.00 92.88 01:50:01 AM all 4.07 0.01 0.85 3.17 0.00 91.90 02:00:01 AM all 4.16 0.01 0.83 1.30 0.00 93.71 02:10:03 AM all 4.42 0.01 0.94 28.86 0.00 65.77 02:20:01 AM all 4.16 0.01 0.92 11.04 0.00 83.88 02:30:01 AM all 3.99 0.00 0.80 5.63 0.00 89.57 02:40:01 AM all 4.00 0.00 0.93 3.09 0.00 91.98 02:50:01 AM all 4.03 0.00 0.97 2.20 0.00 92.79 03:00:01 AM all 3.72 0.01 0.83 1.00 0.00 94.45 03:10:04 AM all 4.68 3.48 1.24 42.44 0.00 48.15 03:20:03 AM all 5.26 0.28 1.53 61.01 0.00 31.92 03:30:01 AM all 4.36 0.00 0.91 15.64 0.00 79.08 03:40:02 AM all 6.97 0.01 0.99 19.77 0.00 72.26 03:50:01 AM all 4.05 0.00 0.90 11.24 0.00 83.81 04:00:01 AM all 3.94 0.00 0.95 3.80 0.00 91.31 04:10:01 AM all 4.26 0.00 1.08 23.94 0.00 70.71 04:20:01 AM all 4.42 0.00 1.03 11.86 0.00 82.68 04:30:01 AM all 3.95 0.01 0.85 2.86 0.00 92.32 04:40:01 AM all 3.88 0.01 0.80 2.25 0.00 93.06 04:50:01 AM all 4.06 0.00 1.12 6.39 0.00 88.42 05:00:01 AM all 3.83 0.00 0.77 9.31 0.00 86.08 05:10:17 AM all 1.93 0.00 0.67 94.88 0.00 2.51 05:20:15 AM all 0.64 0.00 0.70 98.67 0.00 0.00 05:30:02 AM all 7.86 0.01 1.24 89.18 0.00 1.70 05:40:01 AM all 3.93 0.00 0.79 19.51 0.00 75.76 05:50:01 AM all 3.58 0.01 0.72 6.91 0.00 88.78 06:00:01 AM all 3.86 0.00 0.79 5.55 0.00 89.79 06:10:01 AM all 4.30 0.01 0.80 9.41 0.00 85.48 06:20:01 AM all 4.84 0.00 0.78 4.08 0.00 90.30 06:30:01 AM all 4.67 0.01 0.78 2.50 0.00 92.05 06:40:01 AM all 3.79 0.01 0.81 1.90 0.00 93.49 06:50:01 AM all 3.80 0.01 0.79 2.20 0.00 93.21 07:00:01 AM all 3.72 0.00 0.77 1.83 0.00 93.68 07:10:01 AM all 3.89 0.00 0.77 3.80 0.00 91.54 07:20:01 AM all 3.75 0.01 0.70 2.28 0.00 93.27 Average: all 4.23 0.10 0.91 18.42 0.00 76.33 --- Server: vps Date: Saturday, 26 October 2013 5:15:37 AEST Notification: vps has lost service Proc-VMware-vmmemctl. followd by Date: Saturday, 26 October 2013 5:17:33 AEST Notification: vps has regained service Proc-VMware-vmmemctl. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
[SLUG] assessing vps performance issues
I'm experiencing serious performance issues on a VPS, as far as I concerned, it's some issues on the actual server not the vps what if anything can I test on the vps to 'prove my suspicions', both past and future ? I have some stuff like cacti and sar already installed last night, htop was showing load average of over 90, according to cacti, it even might have peaked at 100+ around 2am this morning, it's at it's usual of around .15 # sar -s 01:00:00 -e 02:30:00 Linux 2.6.32-358.23.2.el6.x86_64 10/25/2013 _x86_64_(2 CPU) 01:20:21 AM CPU %user %nice %system %iowait%steal %idle 01:32:51 AM all 2.64 0.00 1.07 96.29 0.00 0.00 01:40:04 AM all 3.98 0.00 1.11 94.90 0.00 0.00 01:50:09 AM all 9.72 0.01 1.30 87.94 0.00 1.04 02:00:06 AM all 9.47 0.01 1.72 75.23 0.00 13.57 02:10:07 AM all 4.82 0.01 1.11 92.58 0.00 1.49 02:20:02 AM all 6.54 0.01 1.18 52.34 0.00 39.93 Average:all 6.16 0.01 1.25 83.24 0.00 9.35 -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
On Fri, October 25, 2013 9:55 am, David Bomba wrote: The reporting of CPU consumption is based solely on the processes your VPS is running. You may want to log the processes running during periods of high load, perhaps logrotate or similar cron task is smashing CPU late at night. David, thanks when I saw the high usage, htop reported httpd as biggest user, shutting httpd down made no difference how would an overloaded hd storage manifest itself ? -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
Significant iowait may indicate poor performing storage. Run bonnie++ on your server to give an indication of performance. David Echoman PTY LTD On 25/10/2013, at 10:10 AM, li...@sbt.net.au wrote: On Fri, October 25, 2013 9:55 am, David Bomba wrote: The reporting of CPU consumption is based solely on the processes your VPS is running. You may want to log the processes running during periods of high load, perhaps logrotate or similar cron task is smashing CPU late at night. David, thanks when I saw the high usage, htop reported httpd as biggest user, shutting httpd down made no difference how would an overloaded hd storage manifest itself ? -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
On Fri, October 25, 2013 10:22 am, David Bomba wrote: Significant iowait may indicate poor performing storage. Run bonnie++ on your server to give an indication of performance. thanks, I'll look at that fwiw, the data centre now admitted to 'overloaded storage' that affected several machines -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
Makes sense, If httpd has to wait on storage, then more httpd processes would be started to serve other requests On 25 October 2013 10:27, li...@sbt.net.au wrote: On Fri, October 25, 2013 10:22 am, David Bomba wrote: Significant iowait may indicate poor performing storage. Run bonnie++ on your server to give an indication of performance. thanks, I'll look at that fwiw, the data centre now admitted to 'overloaded storage' that affected several machines -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
I know it's resolved, but look at the 6th column in the original post. 90+% iowait, 90% of the time, the cpu is waiting for IO. now what you should do is run the same command when everything is working normally so you'll know what the output looks like. Then when you have another problem, you can compare with the normal output. -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
On Fri, October 25, 2013 1:36 pm, Michael Chesterton wrote: I know it's resolved, but look at the 6th column in the original post. 90+% iowait, 90% of the time, the cpu is waiting for IO. Michael, thanks I hope it's resolved 'permanently', in the last 48 hours it happened twice, last might from 11pm till about 2:30am, day before from 5am till 9am in the last few weeks, my cacti 'load average' crept up, which wasn't really expected, as there was no real changes to server, nor was usage increase observed (possibly, though unlikely) so perhaps overload of disk storage caused VPS load average increase creep ? now what you should do is run the same command when everything is working normally so you'll know what the output looks like. Then when you have another problem, you can compare with the normal output. -- sar -s Linux 2.6.32-358.23.2.el6.x86_64 10/25/2013 _x86_64_(2 CPU) 08:00:01 AM CPU %user %nice %system %iowait%steal %idle 08:10:02 AM all 4.47 0.01 1.00 2.57 0.00 91.95 08:20:02 AM all 4.26 0.01 0.97 4.95 0.00 89.81 08:30:01 AM all 4.00 0.00 0.88 1.54 0.00 93.57 08:40:01 AM all 4.38 0.01 0.87 1.98 0.00 92.76 08:50:01 AM all 4.13 0.01 0.96 5.63 0.00 89.28 09:00:01 AM all 4.71 0.01 1.02 1.19 0.00 93.07 09:10:01 AM all 4.51 0.01 0.98 3.25 0.00 91.26 09:20:01 AM all 3.99 0.01 0.94 2.62 0.00 92.44 09:30:01 AM all 3.70 0.01 0.93 1.25 0.00 94.11 09:40:02 AM all 3.73 0.00 0.83 2.10 0.00 93.35 09:50:01 AM all 4.22 0.01 0.96 2.51 0.00 92.30 10:00:01 AM all 4.53 0.01 1.01 2.37 0.00 92.08 10:10:01 AM all 4.37 0.01 1.10 5.13 0.00 89.40 10:20:02 AM all 4.70 0.01 1.15 6.29 0.00 87.85 10:30:01 AM all 4.35 0.01 1.07 5.80 0.00 88.78 10:40:01 AM all 4.68 0.01 1.18 5.92 0.00 88.21 10:50:01 AM all 4.31 0.01 1.19 11.61 0.00 82.88 11:00:02 AM all 4.97 0.01 1.26 9.37 0.00 84.39 11:10:02 AM all 5.19 0.01 1.27 14.86 0.00 78.68 11:20:01 AM all 4.75 0.01 0.99 4.16 0.00 90.09 11:30:01 AM all 4.54 0.01 0.93 1.32 0.00 93.20 11:40:01 AM all 3.88 0.01 0.84 0.85 0.00 94.42 11:50:01 AM all 4.37 0.01 0.96 2.71 0.00 91.96 12:00:01 PM all 5.15 0.01 1.19 2.46 0.00 91.19 12:10:01 PM all 5.44 0.01 1.23 3.18 0.00 90.14 12:20:01 PM all 5.03 0.02 1.28 3.04 0.00 90.63 12:30:01 PM all 5.39 0.01 1.34 2.84 0.00 90.42 12:40:01 PM all 5.18 0.01 1.24 3.18 0.00 90.39 12:50:01 PM all 5.21 0.01 1.18 2.21 0.00 91.39 01:00:01 PM all 5.40 0.01 1.18 1.61 0.00 91.80 01:10:01 PM all 5.25 0.01 1.11 3.16 0.00 90.47 01:20:01 PM all 5.83 0.01 1.16 2.77 0.00 90.23 01:30:01 PM all 5.01 0.01 1.13 0.97 0.00 92.88 01:40:01 PM all 5.32 0.01 1.04 0.79 0.00 92.85 01:50:01 PM all 4.58 0.01 0.96 1.60 0.00 92.84 02:00:01 PM all 4.73 0.01 0.95 2.62 0.00 91.69 02:10:01 PM all 4.93 0.01 1.04 3.05 0.00 90.97 02:20:01 PM all 4.39 0.01 0.92 2.05 0.00 92.63 02:30:01 PM all 4.46 0.01 0.91 3.06 0.00 91.55 02:40:01 PM all 4.78 0.01 1.15 5.80 0.00 88.27 02:50:02 PM all 5.19 0.01 1.11 3.47 0.00 90.23 03:00:01 PM all 4.24 0.01 0.98 2.86 0.00 91.91 03:10:01 PM all 5.91 0.01 1.21 4.25 0.00 88.62 Average: all 4.70 0.01 1.06 3.60 0.00 90.63 -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
Re: [SLUG] assessing vps performance issues
as well as 'sar' logging I'm currently charting with cacti: traffic, mem usage, disk space, load, users, ping latency, process count; as well as, apache, mysql, dovecot, postfix, icmp what else should I be logging or charting ? -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html