8

My question comes down to, should multiple magento cron:run -vvv processes always be running and hitting MySql constantly.

I am setting up Magento 2.2.1 thru Google Cloud and I have the 3 standard cron jobs that were pre-setup thru Google's 1 click install of Magento.

*/1 * * * * /opt/bitnami/php/bin/php /opt/bitnami/apps/magento/htdocs/bin/magento cron:run -vvv 2>&1

*/1 * * * * /opt/bitnami/php/bin/php /opt/bitnami/apps/magento/htdocs/update/cron.php 2>&1

*/1 * * * * /opt/bitnami/php/bin/php /opt/bitnami/apps/magento/htdocs/bin/magento setup:cron:run -vvv 2>&1

Looking at top -c there is always 2 php.bin processes running, which are hitting MySql constantly and causing it to use around 50% - 70% CPU all the time. Here is a snapshot of what it normally looks like.

 PID USER      PR  NI    VIRT    RES    SHR  S   %CPU   %MEM 
19327 mysql     20   0 3872884 332876  19172 S  60.8  3.4 332:42.45 /opt/bitnami/mysql/bin/mysqld.bin --defaults-file=/opt/bitnami/mysql/my.cnf --basedir=/opt/bitnami+
26458 bitnami   20   0  679516 476444  64492 S  24.6  4.9   0:24.85 /opt/bitnami/php/bin/php.bin /opt/bitnami/apps/magento/htdocs/bin/magento cron:run -vvv
26415 bitnami   20   0  677532 475672  64588 R  23.6  4.9   1:36.11 /opt/bitnami/php/bin/php.bin /opt/bitnami/apps/magento/htdocs/bin/magento cron:run -vvv

I've also changed the crons to run every 5 minutes, instead of the default every minute but the behavior stays the same.

My latest change was alternating every 7 minutes and 8 minutes with the 2 cron:run jobs starting 3 and 4 minutes apart, and with that only 1 cron job is running at a time with 30% - 40% CPU from MySQL.

My site also has no traffic right now because I haven't launched it yet. Is this behavior normal from Magento since there is nothing going on with the site? I let it sit for 12 hours without doing anything at all and when i look at top the cron is still running and hammering MySQL.

UPDATE: It's now clear the problem is just first cron:run process that is causing problems. I changed the 2nd and 3rd items back to every minute and left the first at 8 minutes and there is only a single running cron:run process at a time. From the comment below it could be an issue with Bitnami Magento installations, but this is my first experience with Magento so I don't know if this is expected behavior (I really hope it's not).

  • I'm having similar issues, with a Bitnami-installation. As of now, I'm not able to figure out why, but when I look on the CPU utilization in the dashboard of GCE, I can see that it started out on a few percentages of the CPU some almost 30 days ago. Now, I'm maxing it out and have added 4 x RAM and 4 x number of vCPUs to keep the server alive. Instead of top, I used htop. With it I see that I've got more than ten lines with magento cron:run -vvv. Some has been live for several minutes. I will try to find out why the cron isn't running as expected. – user5198077 Nov 12 '17 at 14:33
  • I get the same behavior when I had the cron set the default every 1 minute. I changed it to run every 7 and 8 minutes as i described in my question and it only spawns 1 process, but it still seems like it's doing far too much. Sound like maybe a bitnami magento problem possibly. – codesmaller Nov 12 '17 at 19:20
  • Yes, changing to every 7 minutes made my server happy. Went from 300%+ cpu utilization and 8 GB of RAM down to 40 % (for one of MySQL paid) and 1.7 GB of RAM. Will investigate how I can turn full logging of Cron on. – user5198077 Nov 12 '17 at 19:38
  • Also the -vvv is maximum verbose logging so arg can be removed from crontab. – codesmaller Nov 12 '17 at 19:40
  • I had a look on the Disk Operations (I/O) chart. When the server appears to be "working", the read operations are close to zero, while the writes are quite high. What happens when the server goes down (or stops responding) is that the read operations passes the writes. Comparing my not so healthy 2.2.0 Magento Bitnami installation with another BItnami (I think still on the 2.1.6 or 2.1.8), the read / write are at very low levels, less than 2, while for the not-so-healthy (but now working) the reads are below 2 but writes often goes beyond 1000! :O – user5198077 Nov 12 '17 at 22:24
6

Providing at least a temporary work around to the problem, to avoid hammering your MySQL server. Git issue describing the problem

At least part of the problem comes down to the cron_schedule table. I'd recommend doing a select count(*) from cron_schedule; and if that returns more than a few hundred, then you have the issue. For me, that query returned 208,046 on a server that has been running for only a few weeks.

If you have the problem then run this query to delete everything in that table, except the recent rows delete from cron_schedule where scheduled_at < date_sub(now(), interval 1 hour);

Then run the count query again and it should be waaaay lower. Mine went from 208k to 252.

After running that query I set all 3 standard crons back to the standard once per minute, and like magic they all 3 run almost instantly. Back to normal as far as i can tell.

In the github issue, another user suggests adding that query to crontab to prevent the table from growing again.

0 * * * * <path_to_mysql_bin_dir>/mysql <magento_db_name> -e "delete from cron_schedule where scheduled_at < date_sub(now(), interval 1 hour)"

The last response from the Magento team on the issue was in September saying they cannot reproduce the issue, but a number of users have followed up saying they experience the same thing, including myself. So hopefully they fix this so this hack can be removed.

EDIT: To use that command on the crontab you'll need to pass your credentials to MySQL in some way. See my comment below if you're on a dedicated VM or server and want to put your user/pass on crontab, otherwise you'll need to set up a MySQL credentials file. And you can use which mysql to find the path to your mysql binary directory.

  • Your answer worked for me, but adding in crontab, with mysql magento-db -e "query", where magento-db is the database name (in my case, it is bitnami_magento), does it also work when there is a password for the database? I usually enter the database like mysql -u somename -p and then enter the password in a prompt. – user5198077 Nov 14 '17 at 23:24
  • Using your preferred search engine it will be easy to find a couple solutions for that; I left that part off for the user/pass info but I'll post it, after i give the disclaimer: Don't do this on shared hosting because it will be visible to others with top -c, among other things. In that case look for how to use a MySQL credentials file, .mycnf or something like that. This is what i have */15 * * * * <path_to_mysql_binary_dir>/mysql -u<sql_user> -p'<sql_user_pass>' <database_name> -e "delete from cron_schedule where scheduled_at < date_sub(now(), interval 1 hour)"; – codesmaller Nov 15 '17 at 3:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.